Morphological preview effects in English are restricted to ...
Transcript of Morphological preview effects in English are restricted to ...
Morphological preview effects in English are restricted to suffixed words
Kelly M. Dann, Aaron Veldre, & Sally Andrews
The University of Sydney, School of Psychology
Short title: Morphology in the parafovea
Contact details
Kelly M. Dann
School of Psychology, The University of Sydney
Sydney, NSW 2006
Australia
© 2021, American Psychological Association. This paper is not the copy of record and may not exactly replicate the final, authoritative version of the article. Please do not copy or cite without authors' permission. The final article will be available, upon publication, via its DOI: 10.1037/xlm0001029
1
Abstract
Much of the evidence for morphological decomposition accounts of complex word
identification has relied on the masked-priming paradigm. However, morphologically
complex words are typically encountered in sentence contexts and processing begins before a
word is fixated, when it is in the parafovea. To evaluate whether the single word-
identification data generalize to natural reading, Experiment 1 investigated the contribution of
morphological structure to the very earliest stages of lexical processing indexed by preview
effects during sentence reading in the gaze-contingent boundary paradigm. Preview
conditions systematically assessed the impact of prefixed and suffixed nonword previews that
manipulated stem and affix overlap, and affix status, against an orthographically legal control
baseline. Initial fixations on suffixed target words showed a preview benefit from nonwords
that combined the target stem with a legitimate affix, but not with a non-affix, while prefixed
targets only benefited from an identical preview. When presented in a masked prime lexical
decision task in Experiment 2, the same stimuli yielded equivalent stem priming from
suffixed and prefixed primes regardless of affix status, consistent with previous masked
priming studies using similar nonword primes. The early effects of morphological structure
selectively observed on parafoveal processing of suffixed words are inconsistent with recent
non-morphological, position-invariant accounts of embedded stem activation. These results
provide the first evidence of morphological parafoveal processing in English and contribute
to recent evidence that readers extract a higher level of information from the parafovea during
natural reading than was previously assumed.
Keywords
Reading, eye movements, morphology, parafoveal processing, masked priming.
2
Morphemes are the smallest meaning-bearing units in language, e.g., a complex word
such as unbreakable can be decomposed into three component morphemes: the prefix un, the
stem break, and the suffix able. A central question in language research concerns the role that
these component morphemes play in lexical retrieval of morphologically complex words.
Most theories of morphological processing assume that component morphemes contribute to
identification of complex words but, despite decades of research, questions about precisely
how and when they are extracted and the relative contribution of orthographic and semantic
information to morphological decomposition remain a source of ongoing debate. Theories of
morphological processing of written words have relied principally on evidence from isolated
word identification tasks and priming paradigms (see Amenta & Crepaldi, 2012;
Diependaele, Grainger, & Sandra, 2012, for reviews). However, in normal reading, complex
words are typically encountered in sentence contexts. The central goal of the present research
is to evaluate whether the morphological effects observed in priming tasks generalize to more
naturalistic sentence reading contexts. Before elaborating the rationale for the eye movement
methods we apply to address this issue, we briefly overview recent debates about
morphological processing.
Decomposition or direct stem activation?
The influential full decomposition account of multimorphemic word recognition (Taft
& Foster, 1975) assumes that the stem morpheme of a complex word (e.g., break) functions
as the ‘access code’ to a cluster of morphologically structured representations. Early models
assumed an initial prelexical parsing process that stripped affixes from the complex word to
isolate the stem required for the access process (Taft, 2004; Taft & Forster, 1975). More
recent extensions of the decomposition account adopt an interactive activation framework in
which form-level representations corresponding to both free stem morphemes, and bound
morphemes, such as affixes, mediate access to whole word representations (e.g., Crepaldi,
Rastle, Coltheart & Nickels, 2010; Diependaele, Sandra & Grainger, 2009) or lemma units
3
(Taft, 2015). In both accounts, morphological segmentation is based solely on orthographic
form, and independent of semantic influences, and therefore referred to as morpho-
orthographic decomposition (Rastle, Davis, & New, 2004).
The strongest evidence for morpho-orthographic decomposition comes from masked
priming lexical decision experiments which show equivalent priming between genuinely
morphologically-related derived word primes and their stems (e.g., hunter – HUNT) as that
observed for pseudo-derived pairs (a real stem + a real affix but without a genuine
morphological relationship e.g., corner – CORN), under conditions in which priming does not
occur for word pairs that have the same degree of orthographic overlap, such as turnip –
TURN, but where ip is not a real affix (e.g., Rastle et al., 2004). This evidence that masked
priming depends on the morphological status of the affix has been widely replicated (see
Rastle & Davis, 2008 for a review) and confirmed by more recent research (e.g.,
Beyersmann, Ziegler et al., 2016) that tightly controlled semantic and morphological factors
claimed to confound interpretation of some of the evidence for morpho-orthographic
decomposition (e.g., Baayen, Milin, Durdevic, Hendrix, & Marelli, 2011). Converging
evidence for the morpho-orthographic account also derives from other methodologies
including masked transposed-letter priming (Beyersmann, Castles, & Coltheart, 2011;
Beyersmann, Duñabeitia, Carreiras, Coltheart, & Castles, 2013) and event-related potentials
(Beyersmann, Iakimova, Ziegler, & Cole, 2014).
The generality of this signature of morphological decomposition has been challenged
by recent evidence that masked priming from nonword primes is insensitive to the
morphological status of the affix. Morris, Porter, Grainger, and Holcomb (2011) reported
equivalent priming from genuinely-derived word primes (flexible- FLEX), pseudo-derived
non-word primes (flexify-FLEX) and non-affixed non-word primes (flexint– FLEX) (see also
Milin, Feldman, Ramscar, Hendrix, & Baayen, 2017). Parallel evidence of equivalent
priming of French words by pseudo-derived and non-affixed non-words (Beyersmann,
4
Casalis, Ziegler, & Grainger, 2015; Beyersmann, Cavalli, Casalis & Cole, 2016) led to the
conclusion that the stem morpheme in a complex word is activated without explicit prior
decomposition.
To account for these data, Grainger and Beyersmann (2017) proposed an edge-aligned
embedded word activation account of morphological decomposition which assumes that
lexical access occurs through direct activation of the stem embedded in a complex word,
without requiring pre-lexical processing to decompose and isolate the stem. Instead, the
whole-word representation of an embedded stem is directly activated by a complex word in
the same way as it would be for the simple form, as long as the embedded word is aligned
with the beginning or end of the word. The absence of an adjacent letter is assumed to
enhance the visual salience of the embedded word and support automatic activation of the
whole-word representation of the stem. To reconcile the lack of effect of affix status for
nonword previews with the previous results obtained with word primes (e.g., stronger
priming for corner – CORN than turnip - TURN), Grainger and Beyersmann (2017) argued
that the whole word-representation of the word prime (e.g., corner, turnip) yields inhibition
that can be counteracted by the presence of a genuine affix in the segmented letter-string
which ‘boosts’ activation of the stem.
Thus, in contrast to full decomposition models, the embedded word activation account
assumes that the initial segmentation of complex words is “a strictly non-morphological
process” (p. 289) that relies on representations of whole-word forms rather than specialized
morphological representations or affix-stripping mechanisms. However, like full
decomposition approaches, it assumes that skilled readers extract stems via an automatic
“semantically ‘blind’ morpho-orthographic parsing mechanism” (Grainger & Beyersmann,
2017, p. 308) which is assumed to develop at higher levels of reading proficiency through
reading experience and exposure to morphologically complex words which enables
orthographic representations to be established for both whole word forms and affixes (e.g.,
5
Beyersmann, Grainger et al., 2015; Hasenacker et al., 2016). The validity of pure
orthographic accounts of morphological decomposition has been challenged by evidence that
masked priming can be influenced by the semantic transparency of the relationship between
the prime and target (e.g., Feldman, O’Connor, & Moscoso del Prado Martin, 2009; Jared,
Jouralev, & Joanisse, 2017). Morphemes lie at the intersection between orthography and
semantics, so early effects of morphological structure are consistent with a range of
theoretical accounts that argue for the “simultaneous engagement of all sources of
information – formal or meaning-related – in the task of identifying a word in print”
(Schmidtke, Matsuki, & Kuperman, 2017, p. 1810), and challenge the claim that morpho-
orthographic decomposition is ‘semantically blind’.
The present research was not designed to directly assess the relative contribution of
orthographic and semantic processing to decomposition, although it may contribute evidence
relevant to that issue. However, the central goal was to evaluate the generality and timecourse
of morphological decomposition effects – whatever their source – by investigating whether
they generalize beyond the masked priming tasks that have dominated the literature on
morphological processing to eye-movement indices during sentence reading. To provide
novel evidence about the nature and timecourse of decomposition, we investigated whether it
influences the earliest stages of lexical processing, which begin when the word is in the
parafovea - before it is directly fixated. In this context, the critical evidence for a
morphological decomposition process is that it is restricted to stimuli that are ‘exhaustively
decomposable’ into morphemic units (Milin et al., 2017). That is, morphological
decomposition of complex nonwords will be specific to stimuli that combine a word stem
with a genuine affix.
Prefixed vs. suffixed words
Further insight into the decomposition process is provided by a comparison of
prefixed and suffixed words. Both the full decomposition and the edge-aligned embedded
6
word activation accounts assign priority to the stem due to its status as a lexical unit
independent of its functional role in a complex word. In the decomposition accounts, affixes
are parsed pre-lexically to reveal the stem access code; in the edge-aligned account the stem
is directly activated, and any ‘boost’ to activation due to morphological structure of the letter
string occurs post-lexically. Both accounts therefore imply that the processing of prefixed and
suffixed words should not differ.
However, there are a number of functional differences between prefixes and suffixes
that may influence how they are processed. Suffixed words include both inflected and derived
forms which can change the stem both semantically and syntactically, while prefixes are
typically derived forms, so change the meaning of the word only. Orthographically, suffixes
can also change the form of a stem (e.g. dropping the e in composing), while prefixes
preserve the entire stem. Differences in the time-course of processing may also arise due to
the left-to-right parsing that is assumed in many models of English word identification (e.g.,
Coltheart, Rastle, Perry, Langdon & Ziegler, 2001; Davis, 2010). The basic orthographic
syllabic structure (BOSS) model predicts the stem access code will be extracted earlier in a
suffixed word such as payment, than in a prefixed word such as prepay (Taft, 1979). On the
other hand, the affix of a prefixed word is available earlier for affix-stripping. Furthermore,
during reading of connected text, the distance between the point of fixation and the stem/affix
of an upcoming word will differ for prefixed versus suffixed words, which may influence
when morphological information becomes available.
Relatively few studies have directly compared prefixed and suffixed forms. Early
work in French by Cole, Beauvillain, and Segui (1989) found that lexical decision latencies
for suffixed words displayed stem frequency effects, consistent with morphological
decomposition and lexical access via the stem morpheme, while prefixed words showed
surface frequency effects, consistent with whole-word processing. They concluded that
morphological decomposition is not obligatory, and that differences may be due to the
7
interaction between the sequential organization of prefixes and suffixes, and left-to-right
parsing. For suffixed words the stem is available first, so rapid on-line processing is best
achieved by access via the stem, but the stem of a prefixed word can only be determined once
the prefix is identified, and therefore lexical access often proceeds via the whole-word form.
In contrast to this conclusion, Feldman and Larabee (2001) reported that lexical decision
responses to suffixed targets (e.g., payment) showed equivalent facilitation from prefixed and
suffixed morphological relatives (e.g., prepay vs payable) in both short lag (250 ms) and long
lag (average of 10 items) visual priming paradigms suggesting that the benefit of a shared
stem morpheme did not depend on the relative order of the stem and affix.
This conclusion is supported by recent research that used the masked priming
paradigm to directly compare processing of prefixed and suffixed forms. Beyersmann,
Cavalli et al. (2016) found equivalent priming of French target words by genuinely prefixed
and suffixed non-word primes (preamour – AMOUR, amouresse – AMOUR) and non-affixed
primes (brosamour – AMOUR, amourugne – AMOUR) relative to an unrelated control
condition. Similar findings for prefixed and suffixed words in English were recently reported
by Heathcote et al. (2018): both genuinely-affixed (subcheap – CHEAP, cheapize - CHEAP)
and non-affixed (blacheap – CHEAP, cheapstry - CHEAP) nonword primes yielded
significantly greater priming than an unrelated condition (miscall – CHEAP, idealism -
CHEAP). Heathcote et al. (2018) interpreted the results as supporting Grainger and
Beyersmann’s (2017) claim that position-invariant embedded word activation is independent
of morphological structure. However, in contrast to Beyersmann, Cavalli, et al. (2016),
Heathcote et al. (2018) observed a graded pattern of priming: the affixed condition yielded
significantly more priming than the non-affixed condition. This was attributed to the use of
‘interpretable’ affixes (e.g., subcheap, cheapize) which may facilitate post-lexical “attempts
to ‘re-combine’ stem and affix … to compute a meaning for the novel input string” (p. 8) and
boost activation of the embedded stem. However, they acknowledged that the graded priming
8
effects were also compatible with evidence of early semantic influences on decomposition
(e.g., Feldman et al., 2009; Jared et al., 2017), raising questions about whether the activation
of embedded word stems is the “strictly non-morphological process” proposed by Grainger
and Beyersmann (2017, p. 289).
Using eye movements to assess morphological processing
The single word identification and masked priming tasks that have provided the
majority of evidence about morphological processing provide limited insight into normal
reading, where complex words are typically encountered in sentence contexts. Studies of eye
movements during natural sentence reading provide greater ecological validity. They also
yield more direct evidence about the time course of morphological effects. Eye-movement
studies have demonstrated that lexical processing begins before a word is directly fixated -
when it is in the parafoveal area of vision. The influence of parafoveal processing in normal
sentence reading is typically assessed using the gaze-contingent boundary paradigm (Rayner,
1975), which allows for experimental manipulation of previews which are only available for
processing in the parafovea. When the reader’s eyes cross an invisible boundary to the left of
the target word, the preview is replaced by the correct form of the target word. Readers
therefore always fixate the correct target word, and are typically unaware of the display
change from the preview to the target because it occurs during a saccade. Nevertheless,
fixations on the target word are shorter when it replaces a valid or related preview than a
different, invalid preview, a phenomenon known as the parafoveal preview effect (see
Schotter, Angele, & Rayner, 2012; Vasilev & Angele, 2017, for reviews). The boundary
paradigm can therefore be used to investigate manipulations of preview-target similarity like
those investigated in single word masked priming tasks.
Until recently, parafoveal preview effects in English were thought to be limited to
processing of low-level orthographic (e.g., Inhoff & Tousman, 1990) and phonological (e.g.,
Pollatsek, Lesch, Morris, & Rayner, 1992) information, and it was assumed that higher-level
9
information was not processed parafoveally. However, recent studies have demonstrated that
semantic and syntactic attributes of previews are processed in the parafovea (Schotter & Jia,
2016; Veldre & Andrews, 2016, 2017, 2018a, 2018b, 2018c; Veldre, Reichle, Wong, &
Andrews, 2020; see Andrews & Veldre, 2019, for a review). These findings renew interest in
the question of whether higher-level morphological information may also be extracted from
the parafovea.
Despite a large body of eye-movement evidence on the processing of compound
words (e.g., Andrews, Miller, & Rayner, 2004; Bertram & Hyönä, 2003; Hyönä & Pollatsek,
1998; Juhasz, 2008), relatively few studies have used eye movements to assess reading of
other morphologically-complex words. Beauvillain (1996) reported differences in the
timecourse of processing prefixed and suffixed French words that paralleled Cole et al.’s
(1989) lexical decision data: Suffixed words displayed stem frequency effects early in the
eye-movement record, suggesting that lexical access proceeds via the stem, but prefixed
words showed a later effect that only emerged on refixations, suggesting that lexical access
relied on the whole-word form.
More recent research in German has used the fast-priming paradigm (Sereno &
Rayner, 1992) to extend the results of single-word studies to sentence reading using eye-
movement recording. This procedure presents a brief prime after readers fixate on the target
location, which is then replaced with the target word. In contrast to the boundary paradigm,
the display change occurs in a fixated location so readers are consciously aware of it.
Consistent with the single-word masked-priming results for nonword primes (Beyersmann,
Cavalli et al., 2016), Mousikou and Schroeder (2019) found equivalent priming from
genuinely-affixed and non-affixed nonword primes, and no differences between prefixed and
suffixed forms. These results provide convergent evidence for the edge-aligned account
(Grainger & Beyersmann, 2017) in a more ecologically valid task. However, like masked
priming, fast priming only assesses early processing of a fixated target word. During normal
10
sentence reading initial processing starts before a word is fixated. To investigate whether
morphological decomposition affects the earliest stages of lexical retrieval therefore requires
assessment of parafoveal processing.
Morphological processing in the parafovea. Evidence of parafoveal processing of
morphological information has been found cross-linguistically. In Hebrew, derived-form
previews which share the same root as the target have been found to yield a benefit on early
measures (first fixation and gaze duration) compared to an orthographic control (Deutsch,
Frost, Pelleg, Pollatsek, & Rayner, 2003; Deutsch, Frost, Pollatsek, & Rayner, 2005). In
Russian, inflected-form previews which share the same root as the target produced a preview
cost on early measures compared to an identical preview (Stoops & Christianson, 2017,
2019). Similarly, Korean previews with an incorrect morphosyntactic case-marker yielded a
preview cost (Kim, Radach, & Vorstius, 2012). In addition to preview effects, morphological
structure has also been shown to influence saccade-targeting. Although the initial landing
position is typically near the center of a word (i.e., preferred viewing location; Rayner, 1979),
studies in Uighur (Yan et al., 2014) and in Finnish (Hyönä, Yan, & Vainio, 2018) have
demonstrated an effect of morphological complexity: controlling for word length and launch
site, initial landing positions were closer to the beginning of words that contained more
affixes. Similar effect on landing position were observed in a recent comparison of parafoveal
preview effects in suffixed and monomorphemic Finnish words (Hyönä, Heikkilä, Vanio, &
Kliegl, 2021) but the study did not include a manipulation of the legality of the invalid affix
previews to determine whether parafoveal morphological influences were enhanced for
exhaustively decomposable nonword previews.
The only studies of morphological preview effects in English that we are aware of
failed to find evidence of parafoveal morphological processing of prefixed words (Kambe,
2004; Lima, 1987) or compound words (Juhasz, White, Liversedge, & Rayner, 2008). This
may indicate that such effects are limited to morphologically-rich languages like Hebrew,
11
Finnish and Russian. However, no studies to date in English have investigated suffixed
forms. Further, both Kambe (2004) and Lima (1987) assessed preview effects of
morphologically related nonwords that replaced either the affix or the stem with random
consonants (e.g., rehsxc or zvduce as a preview for reduce) relative to baselines of X-mask
placeholders (e.g., xxxxxx) and illegal consonant-strings (cwxyjq). Orthographically atypical
parafoveal previews have been demonstrated to attract attention parafoveally, yielding both
parafoveal-on-foveal disruption (Rayner, 1975; Schotter et al., 2012) and high preview costs
(e.g., Hutzler et al., 2013; Veldre & Andrews, 2018b). Assessing preview effects against
orthographically illegal baselines, and including illegal strings in the morphologically related
conditions, may have reduced sensitivity to Kambe’s and Lima’s morphological
manipulations.
The current study
The aim of the current research is to contribute novel evidence to the debate about
pre-lexical decomposition by investigating morphological influences on word identification
during sentence reading. Experiment 1 assessed the contribution of morphological structure
to the very earliest stages of lexical processing in on-line sentence reading indexed by
parafoveal preview effects in the boundary paradigm. The experiment improved upon past
studies using the gaze-contingent boundary paradigm to investigate morphological processing
in English (Kambe, 2004; Lima, 1987) by directly comparing prefixed and suffixed words.
Paralleling recent masked priming studies (Beyersmann, Cavalli et al., 2016; Heathcote et al.,
2018), the preview conditions assessed the impact of morphologically structured nonwords
against an orthographically legal control baseline. As illustrated by the example in Figure 1,
the five preview conditions systematically varied the degree of morphological overlap with
prefixed and suffixed targets by comparing reading times on the same target word when
preceded by nonword previews that were: (1) identical to target, (2) same stem + different
genuine affix, (3) same stem + non-affix, (4) different stem + same affix, and (5) different
12
stem + different affix (control). Greater preview benefit for the genuinely-affixed condition
than the non-affixed condition would support theories that assume early, exhaustive
morphological decomposition. However, evidence of equivalent preview benefits for same-
stem conditions regardless of affix type, and no benefit for a preview that only shared an affix
with the target, would be more compatible with theories that attribute morphological effects
to the lexical activation of the embedded stem.
(Insert Figure 1 here)
Both theoretical frameworks assume that similar morphological processes are applied
to prefixed and suffixed words in single word tasks in which stimuli are centrally fixated.
However, differences may arise in parafoveal processing due to either visual acuity or left-to-
right parsing mechanisms. The stems of suffixed words are closer to the fovea than those of
prefixed words, and processed earlier in the left-to-right sequence, so may therefore be more
likely to be extracted when the word is in the parafovea. Alternatively, if morphological
decomposition depends on some form of pre-lexical ‘affix-stripping’, the greater proximity of
prefixes in parafoveal words might yield stronger early morphological influences for prefixed
than suffixed targets.
To evaluate whether any morphological effects observed in the eye movement data
for Experiment 1 were due to idiosyncratic characteristics of our stimulus materials, we
conducted a masked prime lexical decision task (Experiment 2) using the stimuli from
Experiment 1 to allow direct comparison with the nonword priming effects that underpin
Grainger and Beyersmann’s (2017) edge-aligned account of morphological priming.
Experiment 1
Method
Participants. Fifty-two undergraduate students (Mage = 19.25, SD = 2.08, 61.5%
female) from The University of Sydney completed the experiment for partial course credit.
13
All participants reported normal or corrected-to-normal vision, and that English was either
the first language they learned to read and write, or that it had been learned prior to age 5.
The study was approved by The University of Sydney Human Research Ethics Committee,
and participants provided written consent prior to participating in the study.
Materials and Design. One-hundred prefixed and 100 suffixed target words were
selected from the English Lexicon Project (ELP) database (Balota et al., 2007). All target
words were bimorphemic; i.e. comprised of a stem and one affix, and between 6 and 10
letters long. All targets were derived forms which contained a whole embedded stem,
unchanged by derivation (e.g., education would not meet criteria as educate is not embedded
in its whole form) to allow assessment of embedded-stem priming effects. To ensure all
target words were easily identifiable, only those with mean lexical decision accuracy greater
than 90% in the ELP database were selected. Prefixed and suffixed word lists were matched
as closely as possible on important psycholinguistic variables that may influence word
recognition (see Table 1).
(Insert Table 1 here)
There was no significant difference in ELP lexical decision accuracy for whole-word
(t = 1.86, p = .065) or stem forms (t = 1.12, p = .263), or in log SUBTLEX frequency
(Brysbaert & New, 2009) for whole word forms (t = -1.29, p = .199). However, stem
frequency was higher for prefixed words (t = 3.44, p = .001). Word length was matched for
stems (t < 1), however due to the prefixes for words higher than 90% in ELP accuracy being
shorter than suffixes, suffixed targets were slightly longer (t = -5.13, p < .001). To control for
the differences in the stem frequency and target length of prefixed and suffixed words, both
variables were included as covariates in the analyses. See Supplementary Materials for the
complete list of stimuli.
Previews in the gaze-contingent boundary paradigm must be of the same length as the
target, as a change in length is a salient cue that increases awareness of the display change.
14
Non-word previews requiring different stems and different affixes were therefore created
from stems and affixes of the same length as the target words and of approximately
equivalent frequency. Non-word stem-affix combinations were always legal combinations of
letters i.e., they exist in genuine words. Non-word previews requiring non-affixes were
constructed from the non-affixes used by Heathcote et al. (2018) supplemented by a further
three two-letter non-affixes. It was not possible to match the variety of affixes in the suffixed
target words (25) and the prefixed target words (13). The productivity of the two sets of
affixes, as indexed by morphological family size (Rubin, 1978) was also significantly lower
for the set of suffixes than prefixes (355 vs 647; t = 5.66, p <.001). These differences reflect
the distribution of affixes in the SUBTLEX database and, more generally, in the corpus.
Additional analyses including affix length and productivity as covariates were conducted to
control for these differences.
Target words were embedded in neutral sentence frames of 7-14 words long (Mprefix =
10.48, Msuffix = 10.41, see Supplementary Materials). The target word was always embedded
more than two words from the beginning, but less than two words from the end of the
sentence, and was preceded by a pre-target word that was at least four letters in length. The
sentences were constructed to ensure that the target word was not predictable from the
preceding sentence context (see below). Five counterbalanced lists were created in which
each of the 200 sentences only occurred once, but across lists each sentence appeared in all
preview conditions.
Stimulus validation. The predictability of the target word in the sentences was
assessed in a separate norming study. A group of 21 students from The University of Sydney
who did not participate in the main experiment provided cloze norming data. Participants
were presented with all 200 sentences up to, and including, the pre-target word, and asked to
provide one word that they thought was most likely to come next in the sentence. Results of
the cloze task showed that the mean cloze probability for the target word was extremely low
15
(M = 0.30%; range: 0-14%) and was similar for the prefixed and suffixed conditions (Mprefix =
0.48%; Msuffix = 0.19%). In addition to assessing whole-word predictability, the cloze
responses were also coded for stem and affix predictability, which were both also close to
zero: Mstem= 1.33% (prefixed: 1.62%; suffixed: 1.05%); Maffix=0.20% (prefixed: 0.43%;
suffixed: 0%). Thus, the target words, stems, and affixes were all very low in contextual
predictability.
Apparatus. Eye movements were recorded using a SR Research Eyelink 1000 Plus
with a sampling rate of 2000 Hz. Subjects read sentences displayed on a 21-in. View Sonic
CRT monitor with a refresh rate of 140 Hz. Sentences were displayed as a single line of text
in black monospaced Consolas font on a grey background. Participants were seated 60 cm
from the monitor, at which distance 2.5 characters subtended 10 of visual angle. Head
movements were minimized with a chin and forehead rest. Viewing was binocular, but only
the right eye was recorded.
Procedure. Participants were instructed to read the sentences silently for meaning,
and to answer occasional comprehension questions by pressing a key on a standard keyboard.
The experiment began with three practice trials to familiarize participants with the procedure.
This was followed by the individually randomized presentation of the 200 experimental
sentences, combined with 108 filler sentences.1 Following approximately 33% of sentences, a
two-alternative yes/no comprehension question requiring moderate understanding was
presented. The frequency of questions was matched between the prefixed and suffixed items.
At the beginning of the experiment, and before each block of approximately 60 sentences,
participants’ eye movements were calibrated using a 3-point calibration procedure. Maximum
calibration error accepted was 0.30 of visual angle. At the beginning of each trial, a circular
black fixation point appeared on the screen at the location of the first letter of the sentence as
1 Filler sentences comprised the stimuli of an unrelated experiment, of which 67% contained a display change.
16
a gaze trigger. Participants were instructed to fixate on this point, and once their fixation was
stable, the sentence was displayed. If a stable fixation could not be made, or if calibration
error was above 0.3 degrees of visual angle, a new calibration was performed. After the
experiment, participants were asked if they had noticed anything unusual during the
experiment. If participants reported seeing a display change, they were asked to provide
details of what they had seen, and an estimate of how many times they noticed it. The entire
session was completed within 60 minutes.
Results
Data Preparation. Prior to analysis, fixations below 80 ms within one letter space of
an adjacent fixation were merged with that fixation and remaining individual fixations below
80 ms or above 1000 ms were excluded (4.93% of total fixations). Trials were excluded from
analysis if the participant blinked immediately before or after fixating the target word (5.13%
of trials). Trials were also excluded due to late display changes (occurring more than 10 ms
into a fixation); and early display changes, which occur when participants make a “j-hook”
saccade across the boundary before ultimately fixating to the left of the boundary (13.5% of
trials). Fixation duration outliers were also excluded if they exceeded 1000 ms for gaze
duration (< 1% of trials). After these exclusions, 8502 trials (81.75% of the data) were
available for analysis.
After excluding three participants with comprehension scores more than 2.5 standard
deviations below the mean, average comprehension accuracy for the remaining participants
was 94% (SD = 3.26; range 87% – 99%) indicating participants were reading for meaning.
The majority of participants were not aware of the display change manipulation; however,
three participants were excluded due to reporting seeing the display change more than 15
times.2 Therefore, the data from 46 participants was included in the analyses.
2 A further three participants reported seeing the display change more than 15 times but only noticed previews
that were illegal letter strings, which occurred only in the filler sentences. Their data were therefore retained.
17
Reading Measures. Three reading measures were analyzed: first fixation duration
(FFD) on the target word during first-pass reading, regardless of the number of fixations the
word received; single fixation duration (SFD), the duration of the first fixation on the target
word when only one first-pass fixation was made; and gaze duration (GD), the sum of first-
pass fixation durations on the target word. The means for each measure across preview
conditions for prefixed and suffixed target words are presented in Table 2.
(Insert Table 2 here)
The data were analyzed by (generalized) linear mixed-effects models (LMM) using
the lme4 package (Version 1.1-21; Bates, Mächler, Bolker, & Walker, 2015) in R (Version
3.5.3; R Core Team, 2019). Affix type (prefix versus suffix) was coded as an effect contrast.
The models included the fixed effects of stem frequency and target length as mean-centered
covariates3, and subject and item random intercepts, subject random slopes for the effects of
affix type and preview type, and item random slopes for the effects of preview type. Models
with more complex random-effects structures failed to converge. To control for the impact of
testing multiple dependent measures on the Type 1 error rate (von der Malsburg & Angele,
2017), the Bonferroni correction of dividing the .05 alpha threshold by three - the number of
dependent measures – was applied yielding alpha of 0.0125 (critical t/z value = 2.49). Effects
that reached the conventional .05 criterion used in previous studies of morphological
processing are noted as marginal. With two qualifications noted below, parallel analyses
performed on log-transformed duration measures yielded the same pattern of significant
results to the analyses of raw measures below.
Two analyses were conducted that used different sets of contrasts to assess the effects
of preview type nested under levels of affix type (i.e., prefixed targets and suffixed targets).
3 Additional models that also included affix productivity and affix length as covariates yielded an identical
pattern of effects of morphology.
18
The preview effect contrasts compared each related preview to the unrelated control condition
to assess the overall preview benefit from each morphologically related condition relative to
an unrelated, but orthographically legal baseline which has served as the index of
morphological priming in many masked priming studies (e.g., Beyersmann et al., 2016). The
successive differences contrasts assessed whether morphological structure yielded graded
preview effects by assessing the relative contribution of different morphological units to
parafoveal preview benefit.
Preview effect contrasts. There were no significant main effects of affix type on any
measure (all |t|s>1.51)4.
For suffixed targets, there was a significant identical preview effect on all measures
(all |t|s>3.05) because the target word was read more quickly when the preview was identical
to the target compared to an unrelated preview. For prefixed targets, this effect was
significant on both FFD (b=-18.60, SE=5.21, t=-3.57) and SFD (b=-18.85, SE=6.66, t=-2.83),
and marginally significant on GD (b=-17.29, SE=7.04, t=-2.46).5
Suffixed targets also showed a stem preview effect because readers benefited from a
preview sharing the same stem as the target word paired with a different suffix. This effect
was significant on FFD (b=-16.32, SE=4.70, t=-3.47), SFD (b=-22.89, SE=6.00, t=-3.82), and
marginally significant on GD (b=-16.85, SE=6.82, t=-2.47).6 The stem preview effect was not
significant on any measures for prefixed targets (all |t|s<1.62).
4 The models including affix length and productivity as covariates revealed a significant affix effect on gaze
duration (b=--16.48, SE=5.85, t=-2.82), and a marginally significant effect on first fixation (b=-7.486, SE=3.57,
t=-2.10), both reflecting longer fixations on suffixed than prefixed words. 5 The identical preview effect for prefixed targets was significant in the analysis of log-transformed gaze
duration (b=-0.06, SE=0.02, t=-3.00). 6 The stem preview effect for suffixed targets was significant in the analysis of log-transformed gaze duration
(b=-0.06, SE=0.02, t=-3.10).
19
The only indication of a preview benefit from stems paired with a non-affix was a
marginally significant effect on gaze duration for suffixed targets (b=--14.37, SE=6.83, t=-
2.10). This comparison was not significant on FFD or SFD measures for suffixed words
(|t|s<1.42), or on any measure for prefixed targets (|t|s<1.48).
There was no benefit from a preview that shared a prefix or suffix with the target
relative to an unrelated control preview on any measure (Prefixed targets: ts<1; Suffixed
targets: |t|s<1.74), i.e., there was no affix preview effect.
Successive differences contrasts. For prefixed targets, there was a significant
difference between the identical preview and the same stem + different affix preview on FFD
and SFD (ts>2.95), but not GD (t<1). However, suffixed targets showed this effect only on
GD (b=19.18, SE=6.61, t=2.90), but not on FFD or SFD (|t|s<1.16). The late emergence of an
identical preview benefit on gaze duration for suffixed words was due to refixations: Readers
were more likely to refixate suffixed targets following a same stem + different affix preview
versus an identical preview (b=0.46, SE=0.12, z=3.65). This comparison was not significant
for prefixed targets (z<1).
For suffixed targets, the same stem + different affix versus same stem + non-affix
contrast was marginally significant on FFD (b=9.75, SE=4.40, t=2.21), and significant on
SFD (b=15.53, SE=5.48, t=2.84), but not GD (t<1). The affix versus non-affix contrast was
not significant for prefixed targets on any measure (all ts<1).
The same stem + non-affix preview did not differ from the same affix + different stem
preview on any measure (Prefixed targets: all ts<1; Suffixed targets: all ts<1). The final
contrast in this set duplicated the non-significant affix priming contrast from the preview
effect comparisons (Prefixed targets: ts<1; Suffixed targets: |t|s<1.79).
In summary, the analyses of both overall preview effects and the graded effects
assessed by the successive differences contrasts revealed different patterns for prefixed and
suffixed targets. Suffixed words showed significant benefits on first-pass fixations from
20
previews that paired stems with a genuine affix relative to both unrelated previews and stems
paired with a non-affix that did not occur for prefixed words. In contrast, prefixed words
showed an early benefit from identical previews relative to same stem + different affix
previews that were not evident for suffixed words.
Supplementary analyses. To determine whether these findings were modulated by
oculomotor factors, supplementary analyses were conducted to assess effects of launch
distance and landing position.
Saccade launch distance. Morphological preview effects might be modulated by
visual acuity. Given the relatively long target words, readers’ perceptual spans may only have
included the entire preview stimulus when their previous fixation was close to the boundary.
Furthermore, the observed differences in preview effects for prefixed versus suffixed targets
might reflect the fact that the stem of a prefixed word lies further into the parafovea thereby
making it more difficult to process. Stem preview effects for prefixed words might therefore
only be observed in cases where the pre-boundary fixation is closer to the target word (Kliegl
et al., 2013). To account for this possible source of variance, the distance between the target
and the launch site of the saccade that crossed the boundary was included in LMMs as a
mean-centered, continuous predictor, along with interactions between launch distance and the
preview effects contrasts nested under levels of affix type.
The results of these analyses produced a virtually identical pattern of significant
effects as described in the main analyses.7 In addition, there were significant effects of launch
distance on all measures. On FFD and SFD, fixation durations were shorter following long
incoming saccades (Prefixed targets: both |t|s>4.99; Suffixed targets: both |t|s>2.60). In
contrast, gaze durations were longer following long incoming saccades, and this effect was
7 The only difference was that the overall stem preview effect for suffixed targets on gaze duration, which was
marginally significant in the main analysis of raw gaze duration, was significant in the analysis controlling for
launch distance (b=-17.17, SE=6.81, t=-2.52).
21
significant for suffixed targets (Prefixed targets: b=1.34, SE=0.75, t=1.79; Suffixed targets:
b=2.97, SE=0.82, t=3.61). The change in the direction of the effect between measures
suggests that, after a long incoming saccade, readers were likely to make a short fixation that
was terminated by a corrective refixation to move the eyes to a more optimal viewing
location for word recognition. Such corrections reduced the length of initial fixations, but
extended gaze duration. This interpretation was supported by a significant effect of launch
distance on first-pass refixation probability (Prefixed targets: b=0.13, SE=0.01, z=9.11;
Suffixed targets: b=0.13, SE=0.02, z=8.22).
There were also significant interactions between launch distance and the identical
preview effect on all measures (Prefixed targets: all |t|s>2.75; Suffixed targets: all |t|s>2.51)
because the preview effect was larger at closer saccade launch distances. This finding is
consistent with evidence that visual acuity modulates preview effects (Kliegl et al., 2013).
However, launch distance did not significantly modulate any of the other preview effects
(Prefixed targets: all |t|s<1.35; Suffixed targets: all |t|s<2.01). Thus, there was no evidence
that the absence of stem preview effects for prefixed targets depended on launch distance.
Initial landing position. As reviewed in the introduction, investigations of the
morphologically rich languages of Uighur (Yan et al., 2014) and Finnish (Hyönä et al., 2018,
2021) have suggested that parafoveal identification of a suffix may cause readers to adjust
their saccade targeting to land their eyes closer to the beginning of the word, to process the
stem, than the typical preferred viewing location slightly to the left of the word’s center
(Rayner, 1979). To test whether similar adjustments were evident in the present data, we
conducted an analysis of landing position on the target word with the preview effect contrasts
nested under the fixed effect of affix type, and mean-centered continuous predictors of launch
distance, stem frequency, and target word length.
Initial landing position was significantly closer to the word beginning following long
incoming saccades (b=-0.38, SE=0.01, t=-53.95), and significantly further into the word for
22
long targets (b=0.17, SE=0.03, t=5.61). However, there was no effect of affix type (t<1), and
no effects of preview type on landing position (Prefixed targets: all ts<1; Suffixed targets: all
ts<1). Thus, there was no evidence that readers used morphological information in the
parafovea to guide where they directed their saccades. Such effects may therefore be
restricted to morphologically rich languages (Hyönä et al., 2018, 2021; Yan et al., 2014).
These results also provide no evidence that the different patterns of preview effects for
prefixed versus suffixed targets were due to systematic differences in initial landing position.
Discussion
The present study provides the first evidence of morphological influences on
parafoveal processing in English, but showed that the early effects of morphological structure
indexed by parafoveal preview effects were limited to suffixed words. Prefixed words
showed a preview benefit from an identical preview, but no effects of stem overlap even at
close launch sites. In contrast, suffixed words showed as much benefit on first fixation and
single-fixation duration from a same stem + different affix preview as from an identical valid
preview, suggesting parafoveal extraction of the stem.8 However, in contrast with masked
priming results for single words assessing similar conditions, the stem preview benefit was
restricted to previews containing genuine affixes, and did not extend to stems paired with a
non-affix.
Experiment 2
To evaluate whether this evidence of morphological decomposition of suffixed words
was specific to parafoveal preview effects, the same materials were presented in a masked
8 In response to a reviewer’s query, we tested whether the different identity preview effects for prefixed and
suffixed words might be due to differences in position-specific bigram frequency at the affix boundary. These
comparisons showed that the difference between identical previews and same stem+different affix previews was
significant for suffixed conditions (p<.001) but not for prefixed words (p>0.2) and therefore cannot explain why
the identity preview effect was significant for prefixed but not suffixed words; while the difference between the
bigram frequency for same stem+non-affix and unrelated previews was significant for prefixed previews (p<.01)
but not for suffixed previews (p>0.5) so cannot explain why only suffixed words showed any evidence of a
preview benefit for non-affix previews.
23
prime lexical decision experiment to establish whether they replicated the findings of recent
studies of morphological processing using prefixed and suffixed nonword primes
(Beyersmann, Cavalli et al., 2016; Heathcote et al., 2017; Mouiskou & Schroeder, 2019).
Consistent with the methods used in these studies, the target stimuli were word stems and
matched nonwords.
Method
Participants. Fifty undergraduate students (Mage=20.82, SD=4.81, 64% female) from
The University of Sydney completed the experiment for partial course credit. All participants
began to learn to read and write English by at least age 5 and none had completed Experiment
1. The study was approved by The University of Sydney Human Research Ethics Committee,
and participants provided written consent prior to participating in the study.
Materials and Design. The same non-word previews used for the affixed, non-
affixed and unrelated conditions of Experiment 1 were used as the primes in Experiment 2.
Consistent with the majority of masked prime lexical decision experiments, the targets in
Experiment 2 were the stems of the complex word targets used in Experiment 1. Target stems
corresponded to 90 prefixed and 90 suffixed target words from Experiment 1.9 In addition to
these 180 word targets, 180 non-word targets were selected from a list of legal non-words
included in the English Lexicon Project (ELP) database (Balota et al., 2007). The non-words
differed from the target stems by 2-3 letters (e.g. PURE - NIRE) with a range of letter
substitution positions. Prefixed and suffixed targets were matched as closely as possible on
important psycholinguistic variables. Word length was matched (Prefixed: M=5.24, Suffixed:
M= 5.31, t <1) but, in the reduced list, ELP lexical decision accuracy was slightly higher for
prefixed stems than suffixed stems (0.97 vs. 0.95; t = -2.07, p = .02), and, as in Experiment 1,
9 The total number of target stems used in Experiment 2 was reduced from the 100 prefixed and 100 suffixed
complex target words used in Experiment 1 due to removal of cases where two complex words were derived
from the same stem. Because the targets were always stems and primes were always affixed, there was no
parallel to the Identical and same-affix conditions of Experiment 1.
24
log SUBTLEX frequency was higher for prefixed stems than suffixed stems (M= 3.30 vs 2.88
t = -3.36, p <.001). To control for the differences in stem frequency of prefixed and suffixed
words, it was included as a covariate in the analysis10.
Target words were primed by the affixed, non-affixed and unrelated non-word
previews from Experiment 1. Primes for the non-word targets were created to match the word
conditions by modifying the affixed and non-affixed primes for a given word target (e.g.
enpure, napure for PURE) to create the prime for the corresponding non-word (e.g. ennire,
nanire for NIRE). As in Experiment 1, the unrelated prime for a given target stem
corresponded to the non-word created for a different target. The experiment therefore
comprised a total of 360 trials consisting of 180 target words paired with 90 prefixed and 90
suffixed non-word primes (30 x 3 prime types), and 180 target non-words paired with 90
prefixed and 90 suffixed non-word primes. Prefixed and suffixed prime trials were
intermixed and presented in an individually, quasi-randomized sequence that minimized long
sequences of items from the same condition. Three counterbalanced lists were created to
ensure that participants only saw each target once and that, across participants, all targets
were paired with each of the three prime types. Stimuli are included in Supplementary
Materials.
Procedure. Participants completed the experiment individually or in small groups of
up to three in a quiet, darkened room. Stimuli presentation and recording of data was
controlled by DMDX software (Forster & Forster, 2003). Participants were instructed that
their task on each trial was to decide if the string of uppercase letters presented was a real or a
nonsense word, as quickly and accurately as possible. Responses were made by pressing the
left or right shift keys on the keyboard. The experiment began with 18 practice trials to
familiarize participants with the procedure. Each trial began with the 500 ms presentation of a
10 Additional models including affix length and productivity yielded identical outcomes.
25
forward mask of 14 hash marks followed by the prime in lowercase letters for 50 ms, and
then the target in uppercase letters for 3000 ms, or until the participant responded. After the
experiment, participants were debriefed and asked if they had noticed anything unusual
during the experiment. If participants reported any awareness of the prime, they were asked
to provide details of what they had seen.
Results
The majority of participants did not report awareness of the prime, and those who did
were not able to provide any identifying features of the prime stimulus. Therefore, the data
from all 50 participants was included in the analyses. Prior to analysis, latencies below 200ms
(3 trials) or above 1500ms (18 trials) and trials that timed out (21 trials) were removed. For
the RT analysis, incorrect responses (4.54% of the data) were removed. Mean reaction times
and error rates across prime conditions for prefixed and suffixed target words are presented in
Table 3.
(Insert Table 3 here)
The RT data and error data were analyzed by LMM and GLMM, respectively. Affix
type (prefix versus suffix) was coded as an effect contrast. The effect of prime type was
assessed by two planned contrasts nested under levels of affix type: (1) Stem priming effect:
the average of the related primes vs. the unrelated prime, and (2) Morphological priming
effect: affixed vs. non-affixed primes. To control for differences in stem frequency between
the prefixed and suffixed targets, the models included the fixed effect of stem frequency as a
mean-centered covariate. Models included subject and item random intercepts, and by-subject
and by-item random slopes for priming contrasts. Models with more complex random-effects
structures failed to converge. A parallel analysis of log-transformed RT yielded an identical
pattern of results to the analysis of raw RT reported below.
The RT analysis revealed significant stem priming reflecting faster responses to
26
targets primed by related than unrelated primes (Prefixed: b=12.65, SE=4.93, t=2.57;
Suffixed: b=24.37, SE=4.94, t=4.93). However, there was no difference in the size of the
priming effect between the affixed and non-affixed conditions (Prefixed: t<1; Suffixed:
b=8.62, SE=6.02, t=1.43). The main effect of affix type was not significant (t<1).
The analysis of error rates yielded a similar pattern of results as the RT analysis. The
overall error rate was low, but significantly more errors were made in the unrelated prime
condition than in the related conditions for prefixed targets (b=0.60, SE=0.20, z=3.07) but not
suffixed targets (b=0.29, SE=0.15, z=1.88). There were no differences in error rates between
the affixed and non-affixed conditions for either prefixed or suffixed targets (both zs<1). The
main effect of affix type was not significant (b=-0.25, SE=0.21, z=-1.19).
Discussion
Experiment 2 demonstrated that the stimuli used in Experiment 1 replicated the two
critical features of the pattern of masked priming reported in recent studies using
morphologically complex nonword primes. First, consistent with the findings of Beyersmann,
Cavalli et al. (2016) and Mousikou and Schroeder (2019) in French and German,
respectively, we found equivalent masked priming of word stems by affixed and non-affixed
nonword primes. In contrast with Heathcote et al.’s (2019) study of English, the data did not
show enhanced priming for nonword primes constructed from genuine affixes, consistent
with their suggestion that this boost in priming may be limited to stimuli explicitly
constructed to be semantically interpretable.
Second, consistent with all three studies, the pattern of priming did not differ between
prefixed and suffixed primes. The masked priming manipulations differed from those used to
assess preview effects in Experiment 1 because the target words in Experiment 1 were all
affixed words. This was essential to equate the length of the previews and targets, which is
necessary to limit participants’ awareness of display changes in the boundary paradigm. Stem
morphemes were used targets in Experiment 2 because this is the procedure adopted in
27
Beyersmann, Cavalli et al. (2016) and Heathcote et al.’s (2019) studies. Mousikou and
Schroeder (2019) found identical patterns of masked priming for stem and affixed targets,
and for blocked and mixed presentation of prefixed and suffixed words suggesting that such
procedural variations have little impact on masked priming effects.
Thus, when tested under the conditions of the masked priming task in which the
complex nonwords were directly fixated, our stimulus materials showed the pattern of
priming predicted by the embedded word activation account (Grainger & Beyersmann, 2017).
This suggests that the discrepancies from this pattern observed in Experiment 1 are specific to
parafoveal preview effects during sentence reading rather than arising from idiosyncrasies of
our stimulus materials.
General Discussion
Traditional full decomposition accounts of complex word identification have been
challenged by recent findings of significant and equivalent masked priming by genuinely-
affixed and non-affixed non-word primes for both prefixed and suffixed forms (Beyersmann,
Cavalli, et al., 2016, Heathcote et al., 2018). These results have been interpreted as evidence
that lexical access for morphologically complex words proceeds via direct activation of the
embedded stem, without pre-lexical decomposition. The present research investigated
whether these findings generalize to online sentence reading by assessing morphological
influences on parafoveal processing using the gaze-contingent boundary paradigm. Recent
evidence that parafoveal preview effects are sensitive to semantic and contextual plausibility
(see Andrews & Veldre, 2019, for review) show that substantial lexical processing often
occurs for parafoveal words, before they are directly fixated. If stems are accessed during the
early stages of lexical retrieval, they should yield parafoveal preview effects on early eye
movement measures paralleling the effects observed in single word masked priming tasks.
The eye movement data of Experiment 1 showed that stems were activated during
parafoveal processing. However, they also demonstrated two critical caveats on these effects
28
that have important theoretical implications. First, early stem preview effects on parafoveal
processing were restricted to suffixed words. Second, the effects observed for suffixed words
appeared to reflect morphological structure rather than activation of embedded stems. The
evidence for these conclusions is summarized below.
Overall preview effects assessed against an orthographically legal unrelated nonword
baseline provide the most direct comparison with the masked priming effects assessed in
single word paradigms (e.g., Beyersmann, Cavalli, et al., 2016; Heathcote et al., 2018), and
avoid the problems associated with the illegal baselines used in Kambe (2004) and Lima’s
(1987) studies of parafoveal morphological processing. For both prefixed and suffixed
targets, these comparisons revealed that both early and late fixation measures were
significantly shorter when the target was preceded by a valid identical preview than when the
preview was an unrelated nonword. These effects replicate the standard benefit of parafoveal
preview observed in the boundary paradigm (Schotter et al., 2012).
However, the major finding of Experiment 1 was that morphological influences on
parafoveal preview effects were restricted to suffixed words, particularly for early measures.
The stem preview effect only occurred for suffixed words; prefixed targets showed no
evidence of a benefit from previews that comprised the same stem paired with a different
affix. The differential stem preview effect for prefixed and suffixed words on early fixations
was confirmed by the graded effects revealed by the successive differences contrasts.
Prefixed words showed a significant benefit for identical over same stem + different affix
previews on first fixation duration and single fixation duration but suffixed words showed as
much benefit from previews of the target stem with a different affix as from identical
previews. Critically, suffixed words also showed a significantly stronger preview effect on
single fixation duration when the same stem was paired with a different, legitimate affix
29
rather than a non-affix (e.g. dietion vs. dietnel as previews for dietary). 11 There was no
parallel benefit for a preview in which the target stem was paired with a non-affix relative to
either an unrelated preview or a preview combining a different stem with the target affix. The
data therefore indicate that parafoveal preview effects for suffixed words on initial fixations
were limited to exhaustively decomposable morphologically structured nonword previews.
Thus, the results provide evidence that parafoveal processing is sensitive to
morphological structure, but only for suffixed words. These findings appear to be
incompatible with Grainger and Beyersmann’s (2017) claim that activation of edge-aligned
embedded stems is morphologically blind, and occurs for both prefixed and suffixed words.
Previous studies that assessed parafoveal morphological processing in English using
the boundary paradigm tested only prefixed words and detected no sensitivity to
morphological structure (Kambe, 2004; Lima, 1987). The current results are therefore
consistent with the null findings for prefixed words, but do not support the conclusion that
morphological information is not extracted from the parafovea. The results of Experiment 1
suggest that morphological structure does influence word identification during sentence
reading in English, but only for suffixed words. The data therefore add to recent evidence of
relatively deep parafoveal processing during reading (see Andrews & Veldre, 2019).
These findings contrast with the results obtained for the same stimuli in the single
word masked priming task used in Experiment 2, which replicated the evidence of equivalent
priming from affixed and non-affixed primes for both prefixed and suffixed forms found in
previous masked priming studies using nonword primes (Beyersmann, Cavalli et al., 2016;
Mousikou & Schroeder, 2019). The results of Experiment 1 also differ from Mousikou and
Schroeder’s (2019) fast-priming eye-movement data in German, which showed equivalent,
11 The effect on FFD was significant by the conventional .05 criterion but failed to exceed the Bonferroni-
adjustment recommended by von der Marlsburg and Angele (2017).
30
significant priming by genuinely-affixed and non-affixed non-word primes across both
prefixed and suffixed forms.
The critical difference between the gaze-contingent boundary paradigm used in
Experiment 1 and Mousikou and Schroeder’s eye tracking task is that the fast priming
procedure presented the affixed nonword prime after participants fixated on the location of
the target word, which was occupied by a mask consisting of unrelated prior to its fixation.
Thus, no parafoveal preview of the affixed nonword prime was available, and it was
presented foveally, at the point of readers’ fixation, as it is in the masked priming task. In
contrast, in the boundary paradigm the preview can only be processed parafoveally – it is
replaced with the target word during the saccade before the reader fixates that location. The
boundary paradigm therefore provides unique insight into parafoveal processing of
morphological information that is not tapped by the fast priming task. The fact that Mousikou
and Schroeder’s fast priming results duplicated the findings of masked priming studies
suggests that position-invariant activation of embedded stems may be specific to conditions
in which a brief prime stimulus appears in the same foveal location as the target. Norris and
Kinoshita’s (2008, 2012) Bayesian Reader model proposes that such conditions “trick the
perceptual system into processing primes and targets as a single perceptual object” (p. 450)
and tap generalized mechanisms involved in recovering information from noisy visual input
rather than indexing activation of specialized linguistic representations of morphemic or
lexical constituents shared by the prime and target. Stem morphemes are typically longer and
more salient than affix units so their letters may be more likely to be successfully recovered
from the noisy composite visual input. Consistent with the possibility that morphologically-
blind stem activation effects are specific to spatiotemporal conditions in which the prime and
target are treated as a single object, effects of morphological structure and semantic
transparency typically emerge with prime durations above approximately 60 ms (e.g., Rastle,
Davis, Marslen-Wilson & Tyler, 2000), which allow perceptual segregation of the prime and
31
target.
Under the more natural reading conditions of the boundary paradigm, where lexical
processing can be initiated before a word is fixated, the results of Experiment 1 suggest that
lexical retrieval mechanisms are sensitive to the morphological structure of suffixed forms in
the parafovea, and do not simply reflect activation of the stem morpheme. However,
consistent with Kambe’s (2004) and Lima’s (1987) findings, parafoveal preview effects did
not occur for prefixed words. There are two possible explanations of the absence of
parafoveal morphological decomposition of prefixed words.
First, it may reflect the constraints of visual acuity: embedded stems of prefixed forms
fall further into the parafoveal region, where visual acuity is reduced. This may limit the
depth of parafoveal processing, and account for the absence of stem preview effects for
prefixed words. However, contradicting this possibility, the results revealed significant
effects of parafoveal manipulations of affix legality for suffixed previews, even though these
units fell further from fixation than the onset of the stems of prefixed words, i.e., in regions
where visual acuity is low. Further evidence against the contribution of visual acuity to the
differences between prefixed and suffixed words is provided by the supplementary analyses
of launch distance and landing position. As expected, launch distance had robust effects on
general sensitivity to parafoveal information reflected in larger identical preview effects at
close launch sites (Kliegl et al., 2013). However, there was no evidence that it modulated the
differences between morphologically-related previews, or that it depended on affix type.
Similarly, there was no evidence of the morphological complexity effect on initial landing
position observed in Uighur (Yan et al., 2014) and Finnish (Hyönä et al., 2018, 2021)
indicating that the parafoveal effects of morphological structure detected in Experiment 1
were not due to systematic differences in initial landing positions on prefixed versus suffixed
words. In combination, then, the present results suggest that visual acuity alone is unlikely to
explain the lack of stem preview effects for prefixed words.
32
There is a second, essentially opposite explanation of how visual acuity might account
for the lack of morphological preview effects for prefixed words. In same-stem + different
affix previews of prefixed targets, the changed component (i.e., the prefix) forms the initial
letters of the preview (e.g., inforest/abforest as previews for deforest). The salience and
proximity of these initial letters in the highest acuity parafoveal region may increase the
disruptive effects of the discrepancy between the preview and the target and lead to a preview
cost that outweighs any benefit from morphological similarity between them. This
interpretation is consistent with a large body of eye-movement evidence showing that word-
initial letters facilitate parafoveal processing during reading (e.g., Briihl & Inhoff, 1995;
Inhoff, 1989; White, Johnson, Liversedge, & Rayner, 2008). The disruptive effects of
changing the initial letters may explain why prefixed targets showed a preview effect for
identical relative to same-stem previews on initial fixation duration (e.g., deforest vs. inforest)
that was not evident for suffixed targets. There were no other significant preview effects for
prefixed words, perhaps because stem preview benefits were counteracted by interference
from the different initial letters. The present study was not designed to distinguish between
‘affix-stripping’ (e.g., Taft & Forster, 1975, 1976) and interactive activation accounts of
morpho-orthographic decomposition (e.g., Taft, 2015). However, the fact that greater
parafoveal availability of prefix units appeared to hinder rather than help early morphological
decomposition seems incompatible with a process that depends on pre-lexical affix-stripping.
Early identification and stripping of the prefix should enhance extraction of the stem, even
when the target contains a different affix. The precise predictions of interactive accounts of
morphographic decomposition are more difficult to specify because they depend on the
relative frequency of the form units for stems and affixes.
In contrast to the results for prefixed targets, the results for suffixed words are
consistent with a process that depends on the preview being exhaustively decomposable into
morphemic units. This implies that morphological structure contributed to the preview effect
33
and that it was not simply due to orthographic overlap between the preview and target, which
was equivalent for the genuinely affixed and non-affixed conditions. The selective stem
preview effect observed for suffixed words may arise because the ‘edge-aligned’ stem of
such words falls in the highest acuity in the parafoveal region, potentially facilitating its early
activation. Grainger and Beyersmann’s (2017) account attributes any additional preview
benefit due to the legitimacy of the affix to the ‘secondary role’ of a genuine affix to ‘boost’
stem activation. The ‘secondary role’ is not well-defined in the edge-aligned account, but
Crepaldi, Hemsworth, Davis, and Rastle (2016) suggested that lexical representations for
different types of morphemic units may have different properties. For example, stems appear
to have position-invariant coding, so that the same stem can be accessed in both prefixed and
suffixed forms, while affixes occur in fixed positions, and therefore rely on position-specific
orthographic codes. Evidence that a shared affix alone is sufficient to yield masked priming
effects for both prefixed (Chateau, Knudsen, & Jared, 2002) and suffixed (Crepaldi et al.,
2016) words led Crepaldi et al. (2016) to propose that affixes may play a more active role in
complex word recognition. However, the absence of significant preview effects for the same
affix + different stem conditions for either prefixed or suffixed words in Experiment 1
suggests that affixes are not automatically extracted from the parafovea. Most critically, the
preview effects on the duration of initial fixations on suffixed words depended on the
legitimacy of the suffix in parafoveal vision. Such early influences of affix status, even for
low acuity regions, appear to be inconsistent with the late, secondary role of morphological
structure proposed by Grainger and Beyersmann (2017).
The present evidence of parafoveal effects of morphological structure on eye
movement measures for suffixed words converges with Schmidtke et al.’s (2017) recent
application of survival analysis (Reingold & Sheridan, 2014) to determine the earliest point at
which a range of orthographic, semantic and morphological variables influenced first fixation
duration during sentence reading. Across three sets of eye-movement data they found a very
34
early influence of the whole word frequency of derived suffixed words that was
contemporaneous with, or even preceded, the effect of the frequency of the word’s stem,
suggesting that morphemes and whole words are processed in parallel. This timecourse was
argued to be incompatible with pure orthographic accounts, in which decomposition into
morphemes is a prerequisite for access to complex word forms, but consistent with a range of
models that “allow for even partial orthographic information to activate meanings of complex
words and their morphemes in a parallel rather than sequential way” (Schmidtke et al., 2017,
p. 1810). Although Schmidtke et al.’s analyses did not directly tap parafoveal processing,
they noted that a number of form-related variables exerted an earlier influence on fixation
time than observed in a parallel analysis of single word lexical decision latencies, consistent
with effects of parafoveal preview, and suggested that further research should explore
whether contextual predictability facilitates early extraction of semantic information from
parafoveal words. Such processes may contribute to the very early effects of morphological
structure observed for suffixed words in Experiment 1.
Conclusion
To our knowledge, the current research provides the first evidence of parafoveal
processing of morphological information during reading in English. The results also highlight
the importance of collecting converging data from multiple paradigms to assess the impact
and timecourse of psycholinguistic manipulations on natural reading (see also Feldman, Dale,
& van Rij, 2019). The present findings suggest that morphological structure influences the
processing of complex words in the parafovea, but only for suffixed forms. These findings
complement recent evidence that readers extract high-level lexical information from the
parafovea (see Andrews & Veldre, 2019). Since lexical retrieval processes appear to begin in
the parafovea, early effects of morphological decomposition in reading may be most clearly
detected in measures based on information extracted prior to fixation.
35
Author Note
We thank Curtis Chan for his assistance with data collection. This research was
supported under Australian Research Council's Discovery Projects funding scheme (project
numbers DP160103224 and DP190100719). The data, analysis script, and stimulus materials
from this study are available at osf.io/sf89p
36
References
Amenta, S., & Crepaldi, D. (2012). Morphological Processing as We Know It: An Analytical
Review of Morphological Effects in Visual Word Identification. Frontiers in
Psychology, 3(232). doi: 10.3389/fpsyg.2012.00232
Andrews, S., Miller, B., & Rayner, K. (2004). Eye movements and morphological
segmentation of compound words: There is a mouse in mousetrap. European Journal
of Cognitive Psychology, 16, 285-311.
Andrews, S., & Veldre, A. (2019). What is the most plausible account of the role of
parafoveal processing in reading? Language and Linguistics Compass, 13, e12344.
Balota, D. A., Yap, M. J., Cortese, M. J., Hutchison, K. A., Kessler, B., Loftis, B., . . .
Treiman, R. (2007). The English Lexicon project. Behavior Research Methods, 39(3),
445-459. doi: http://dx.doi.org/10.3758/BF03193014
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models
using lme4. Journal of Statistical Software, 67, 1-48.
Beauvillain, C. (1996). The integration of morphological and whole-word form information
during eye fixations on prefixed and suffixed words. Journal of Memory and
Language, 35(6), 801-820. doi: http://dx.doi.org/10.1006/jmla.1996.0041
Bertram, R., & Hyönä, J. (2003). The length of a complex word modifies the role of
morphological structure: Evidence from eye movements when reading short and long
Finnish compounds. Journal of Memory and Language, 48, 615-634.
Beyersmann, E., Casalis, S., Ziegler, J. C., & Grainger, J. (2015). Language proficiency and
morpho-orthographic segmentation. Psychonomic Bulletin & Review, 22(4), 1054-
1061.
Beyersmann, E., Castles, A., & Coltheart, M. (2011). Early morphological decomposition
during visual word recognition: Evidence from masked transposed-letter priming.
Psychonomic Bulletin & Review, 18(5), 937-942.
37
Beyersmann, E., Cavalli, E., Casalis, S., & Cole, P. (2016). Embedded stem priming effects
in prefixed and suffixed pseudowords. Scientific Studies of Reading, 20(3), 220-230.
Beyersmann, E., Dunabeitia, J. A., Carreiras, M., Coltheart, M., & Castles, A. (2013). Early
morphological decomposition of suffixed words: Masked priming evidence with
transposed-letter nonword primes. Applied Psycholinguistics, 34(5), 869-892.
Beyersmann, E., Iakimova, G., Ziegler, J. C., & Cole, P. (2014). Semantic processing during
morphological priming: An ERP study. Brain Research, 1579, 45-55.
Beyersmann, E., Ziegler, J. C., Castles, A., Coltheart, M., Kezilas, Y., & Grainger, J. (2016).
Morpho-orthographic segmentation without semantics. Psychonomic Bulletin &
Review, 23(2), 533-539.
Beyersmann, E., Ziegler, J. C., & Grainger, J. (2015). Differences in the processing of
prefixes and suffixes revealed by a letter-search task. Scientific Studies of Reading,
19(5), 360-373. doi: http://dx.doi.org/10.1080/10888438.2015.1057824
Briihl, D., & Inhoff, A. W. (1995). Integrating information across fixations during reading:
The use of orthographic bodies and of exterior letters. Journal of Experimental
Psychology: Learning Memory and Cognition, 21, 55-67.
Brysbaert, M., & New, B. (2009). Moving beyond Kučera and Francis: A critical evaluation
of current word frequency norms and the introduction of a new and improved word
frequency measure for American English. Behavior Research Methods, 41(4), 977-
990. doi: 10.3758/brm.41.4.977
Chateau, D., Knudsen, E. V., & Jared, D. (2002). Masked Priming of Prefixes and the
Influence of Spelling–Meaning Consistency. Brain and Language, 81(1), 587-600.
doi: https://doi.org/10.1006/brln.2001.2549
Cole, P., Beauvillain, C., & Segui, J. (1989). On the representation and processing of prefixed
and suffixed derived words: A differential frequency effect. Journal of Memory and
Language, 28(1), 1-13. doi: https://doi.org/10.1016/0749-596X(89)90025-9
38
Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). The DRC model: A
model of visual word recognition and reading aloud. Psychological Review, 108, 204-
258.
Crepaldi, D., Hemsworth, L., Davis, C. J., & Rastle, K. (2016). Masked suffix priming and
morpheme positional constraints. The Quarterly Journal of Experimental Psychology,
69(1), 113-128. doi: 10.1080/17470218.2015.1027713
Crepaldi, D., Rastle, K., Coltheart, M., & Nickels, L. (2010). ‘Fell’ primes ‘fall’, but does
‘bell’ prime ‘ball’? Masked priming with irregularly-inflected primes. Journal of
Memory and Language, 63(1), 83-99. doi:https://doi.org/10.1016/j.jml.2010.03.002
Davis, C.J. (2010). The Spatial Coding Model of visual word identification. Psychological
Review, 117, 713,758.
Deutsch, A., Frost, R., Pelleg, S., Pollatsek, A., & Rayner, K. (2003). Early morphological
effects in reading: Evidence from parafoveal preview benefit in Hebrew.
Psychonomic Bulletin & Review, 10(2), 415-422. doi: 10.3758/bf03196500
Deutsch, A., Frost, R., Pollatsek, A., & Rayner, K. (2005). Morphological parafoveal preview
benefit effects in reading: Evidence from Hebrew. Language and Cognitive
Processes, 20(1-2), 341-371. doi: 10.1080/01690960444000115
Diependaele, K., Grainger, J., & Sandra, D. (2012). Derivational morphology and skilled
reading. The Cambridge handbook of psycholinguistics, 311, 332.
Diependaele, K., Sandra, D., Grainger, J. (2009). Semantic transparency and masked
morphological priming: The case of prefixed words. Memory & Cognition, 37(6),
895-908. doi:10.3758/mc.37.6.895
Feldman, L. B., Dale, R., & van Rij, J. (2019). Lexical and frequency effects on keystroke
timing: Challenges to a lexical search account from a type-to-copy task. Frontiers in
Communication, 4, 17.
39
Feldman, L. B., & Larabee, J. (2001). Morphological facilitation following prefixed but not
suffixed primes: Lexical architecture or modality-specific processes?. Journal of
Experimental Psychology: Human Perception and Performance, 27(3), 680.
Feldman, L. B., O’Connor, P. A., & del Prado Martín, F. M. (2009). Early morphological
processing is morphosemantic and not simply morpho-orthographic: A violation of
form-then-meaning accounts of word recognition. Psychonomic Bulletin &
Review, 16(4), 684-691.
Forster, K.I., Forster, J.C. (2003). DMDX: A Windows display program with millisecond
accuracy. Behavior Research Methods, Instruments, & Computers 35, 116–124.
https://doi.org/10.3758/BF03195503
Grainger, J., & Beyersmann, E. (2017). Chapter Nine - Edge-Aligned Embedded Word
Activation Initiates Morpho-orthographic Segmentation. In B. H. Ross (Ed.),
Psychology of Learning and Motivation (Vol. 67, pp. 285-317): Academic Press.
Hasenäcker, J., Beyersmann, E., & Schroeder, S. (2016). Masked morphological priming in
German-speaking adults and children: Evidence from response time
distributions. Frontiers in Psychology, 7, 929.
Heathcote, L., Nation, K., Castles, A., & Beyersmann, E. (2018). Do ‘blacheap’ and
‘subcheap’ both prime ‘cheap’? An investigation of morphemic status and position in
early visual word processing. Quarterly Journal of Experimental Psychology, 71(8),
1645-1654. doi: 10.1080/17470218.2017.1362704
Hutzler, F., Fuchs, I., Gagl, B., Schuster, S., Richlan, F., Braun, M., & Hawelka, S. (2013).
Parafoveal X-masks interfere with foveal word recognition: evidence from fixation-
related brain potentials. Frontiers in Systems Neuroscience, 7, 33. doi:
10.3389/fnsys.2013.00033
40
Hyönä, J. (2011). Foveal and parafoveal processing during reading The Oxford handbook of
eye movements. (pp. 819-838). New York, NY, US: Oxford University Press.
Hyönä, J., Heikkilä, T. T., Vainio, S., & Kliegl, R. (2021). Parafoveal access to word stem
during reading: An eye movement study. Cognition, 208, 104547.
Hyönä, J., & Pollatsek, A. (1998). Reading Finnish compound words: Eye fixations are
affected by component morphemes. Journal of Experimental Psychology: Human
Perception and Performance, 24, 1612-1627.
Hyönä, J., Yan, M., & Vainio, S. (2018). Morphological structure influences the initial
landing position in words during reading Finnish. 71(1), 122-130. doi:
10.1080/17470218.2016.1267233
Inhoff, A. W. (1989). Parafoveal processing of words and saccade computation during eye
fixations in reading. Journal of Experimental Psychology: Human Perception and
Performance, 15, 544-555.
Inhoff, A. W., & Tousman, S. (1990). Lexical Priming From Partial-Word Previews. Journal
of Experimental Psychology: Learning, Memory, & Cognition, 16(5), 825-836.
Jared, D., Jouravlev, O., & Joanisse, M. F. (2017). The effect of semantic transparency on the
processing of morphologically derived words: Evidence from decision latencies and
event-related potentials. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 43(3), 422.
Juhasz, B. J. (2008). The processing of compound words in English: Effects of word length
on eye movements during reading. Language and Cognitive Processes, 23, 1057-
1088.
Kambe, G. (2004). Parafoveal processing of prefixed words during eye fixations in reading:
Evidence against morphological influences on parafoveal preprocessing. Perception
& Psychophysics, 66(2), 279-292. doi: 10.3758/bf03194879
41
Kim, Y.-S., Radach, R., & Vorstius, C. (2012). Eye movements and parafoveal processing
during reading in Korean. Reading and Writing, 25(5), 1053-1078. doi:
10.1007/s11145-011-9349-0
Kliegl, R., Hohenstein, S., Yan, M., & McDonald, S. A. (2013). How preview space/time
translates into preview cost/benefit for fixation durations during reading. The
Quarterly Journal of Experimental Psychology, 66, 581-600.
Lima, S. D. (1987). Morphological analysis in sentence reading. Journal of Memory and
Language, 26(1), 84-99. doi: http://dx.doi.org/10.1016/0749-596X(87)90064-7
McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus during a fixation
in reading. Perception & Psychophysics, 17(6), 578-586. doi: 10.3758/BF03203972
Milin, P., Feldman, L. B., Ramscar, M., Hendrix, P., & Baayen, R. H. (2017). Discrimination
in lexical decision. PloS one, 12(2), e0171935.
Morris, J., Porter, J. H., Grainger, J., & Holcomb, P. J. (2011). Effects of lexical status and
morphological complexity in masked priming: An ERP study. Language and
Cognitive Processes, 26(4-6), 558-599. doi: 10.1080/01690965.2010.495482
Mousikou, P., & Schroeder, S. (2019). Morphological processing in single-word and sentence
reading. Journal of Experimental Psychology: Learning, Memory, and Cognition,
45(5), 881-903. doi: 10.1037/xlm0000619
Norris, D., & Kinoshita, S. (2008). Perception as evidence accumulation and Bayesian
inference: Insights from masked priming. Journal of Experimental Psychology:
General, 137(3), 434.
Norris, D., & Kinoshita, S. (2012). Reading through a noisy channel: Why there’s nothing
special about the perception of orthography. Psychological Review, 119, 517-545.
Pollatsek, A., Lesch, M., Morris, R. K., & Rayner, K. (1992). Phonological Codes Are Used
in Integrating Information Across Saccades in Word Identification and Reading.
42
Journal of Experimental Psychology: Human Perception & Performance, 18(1), 148-
162.
R Core Team. (2019). R: A language and environment for statistical computing. R
Foundation for Statistical Computing. https://www.R-project.org/
Rastle, K., & Davis, M. H. (2008). Morphological decomposition based on the analysis of
orthography. Language and Cognitive Processes, 23(7-8), 942-971. doi:
10.1080/01690960802069730
Rastle, K., Davis, M. H., Marslen-Wilson, W. D., & Tyler, L. K. (2000). Morphological and
semantic effects in visual word recognition: A time-course study. Language and
cognitive processes, 15(4-5), 507-537.
Rastle, K., Davis, M. H., & New, B. (2004). The broth in my brother’s brothel: Morpho-
orthographic segmentation in visual word recognition. Psychonomic Bulletin &
Review, 11(6), 1090-1098. doi: 10.3758/bf03196742
Rayner, K. (1975). The perceptual span and peripheral cues in reading. Cognitive
Psychology, 7(1), 65-81. doi: https://doi.org/10.1016/0010-0285(75)90005-5
Rayner, K. (1979). Eye Guidance in Reading: Fixation Locations within Words. Perception,
8(1), 21-30. doi: 10.1068/p080021
Reingold, E. M., & Sheridan, H. (2014). Estimating the divergence point: A novel
distributional analysis procedure for determining the onset of the influence of
experimental variables. Frontiers in Psychology, 5, 1432.
Rubin, D. C. (1978). Word—initial and word—final ngram frequencies. Journal of Reading
Behavior, 10(2), 171-183.
Schmidtke, D., Matsuki, K., & Kuperman, V. (2017). Surviving blind decomposition: A
distributional analysis of the time-course of complex word recognition. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 43(11), 1793.
43
Schotter, E. R., Angele, B., & Rayner, K. (2012). Parafoveal processing in reading. Attention,
Perception, & Psychophysics, 74(1), 5-35.
Schotter, E. R., & Jia, A. (2016). Semantic and plausibility preview benefit effects in English:
Evidence from eye movements. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 42(12), 1839-1866.
Sereno, S. C., & Rayner, K. (1992). Fast priming during eye fixations in reading. Journal of
Experimental Psychology: Human Perception and Performance, 18(1), 173-184. doi:
10.1037/0096-1523.18.1.173
Stoops, A., & Christianson, K. (2017). Parafoveal processing of inflectional morphology on
Russian nouns. Journal of Cognitive Psychology, 29(6), 653-669. doi:
http://dx.doi.org/10.1080/20445911.2017.1310109
Taft, M. (1979). Lexical access-via an orthographic code: The basic orthographic syllabic
structure (BOSS). Journal of Verbal Learning and Verbal Behavior, 18(1), 21-39.
doi: https://doi.org/10.1016/S0022-5371(79)90544-9
Taft, M. (2004). Morphological decomposition and the reverse base frequency effect. The
Quarterly Journal of Experimental Psychology A: Human Experimental Psychology,
57A(4), 745-765. doi: http://dx.doi.org/10.1080/02724980343000477
Taft, M. (2015). The nature of lexical representation in visual word recognition The Oxford
handbook of reading (pp. 99-113). New York, NY: Oxford University Press; US.
Taft, M., & Forster, K. I. (1975). Lexical storage and retrieval of prefixed words. Journal of
Verbal Learning and Verbal Behavior, 14(6), 638-647. doi:
https://doi.org/10.1016/S0022-5371(75)80051-X
Taft, M., & Forster, K. I. (1976). Lexical storage and retrieval of polymorphemic and
polysyllabic words. Journal of Verbal Learning & Verbal Behavior, 15(6), 607-620.
doi: http://dx.doi.org/10.1016/0022-5371%2876%2990054-2
44
Taft, M., Hambly, G., & Kinoshita, S. (1986). Visual and auditory recognition of prefixed
words. The Quarterly Journal of Experimental Psychology Section A, 38(3), 351-365.
doi: 10.1080/14640748608401603
Vasilev, M., & Angele, B. (2017). Parafoveal preview effects from word N + 1 and word N +
2 during reading: A critical review and Bayesian meta-analysis. Psychonomic Bulletin
& Review, 24(3), 666-689. doi:10.3758/s13423-016-1147-x
Veldre, A., & Andrews, S. (2016). Is semantic preview benefit due to relatedness or
plausibility? Journal of Experimental Psychology: Human Perception and
Performance, 42(7), 939-952. doi: http://dx.doi.org/10.1037/xhp0000200
Veldre, A., & Andrews, S. (2017). Parafoveal preview benefit in sentence reading:
Independent effects of plausibility and orthographic relatedness. Psychonomic
Bulletin & Review, 24(2), 519-528. doi: http://dx.doi.org/10.3758/s13423-016-1120-8
Veldre, A., & Andrews, S. (2018a). Beyond cloze probability: Parafoveal processing of
semantic and syntactic information during reading. Journal of Memory and Language,
100, 1-17.
Veldre, A., & Andrews, S. (2018b). How does foveal processing difficulty affect parafoveal
processing during reading? Journal of Memory and Language, 103, 74-90.
Veldre, A., & Andrews, S. (2018c). Parafoveal preview effects depend on both preview
plausibility and target predictability. Quarterly Journal of Experimental Psychology,
71, 64-74.
Veldre, A., Reichle, E. D., Wong, R., & Andrews, S. (2020). The effect of contextual
plausibility on word skipping during reading. Cognition, 197, 104184.
Von der Malsburg, T., & Angele, B. (2017). False positives and other statistical errors in
standard analyses of eye movements in reading. Journal of memory and language, 94,
119-133.
45
White, S. J., Johnson, R. L., Liversedge, S. P., & Rayner, K. (2008). Eye movements when
reading transposed text: The importance of word-beginning letters. Journal of
Experimental Psychology: Human Perception and Performance, 34, 1261-1276.
Yan, M., Zhou, W., Shu, H., Yusupu, R., Miao, D., Krugel, A., & Kliegl, R. (2014). Eye
movements guided by morphological structure: Evidence from the Uighur language.
Cognition, 132(2), 181-215. doi: http://dx.doi.org/10.1016/j.cognition.2014.03.008
46
Table 1
Mean word length (letters), log frequency, and lexical decision task accuracy of target words
Characteristic Target word Stem
Prefixed Suffixed Prefixed Suffixed
Length 7.69 (1.32) 8.57 (1.10) 5.28 (1.26) 5.33 (1.12)
Frequency 1.88 (0.64) 2.01 (0.74) 3.25 (0.88) 2.90 (0.77)
ELP Accuracy 0.97 (0.03) 0.96 (0.03) 0.96 (0.07) 0.95 (0.06)
Note. All reported values taken from the English Lexicon Project (Balota et al., 2007).
47
Table 2
Mean (and standard error) reading measures for prefixed and suffixed target words across
conditions
Preview condition
Measure Identical Same
stem +
different
affix
Same stem
+ non-affix
Different
stem +
same affix
Unrelated
control
Prefixed target e.g. mistrust pretrust whitrust misnomad outnomad
First fixation duration (ms) 248 (6) 263 (6) 260 (6) 265 (6) 267 (6)
Single fixation duration (ms) 257 (8) 276 (6) 275 (8) 276 (6) 279 (7)
Gaze duration (ms) 315 (10) 321 (8) 320 (9) 327 (8) 331 (10)
Refixation probability .30 (.03) .27 (.03) .26 (.03) .30 (.03) .30 (.03)
Suffixed target e.g. stressful stressary stressard engageful engageous
First fixation duration (ms) 259 (6) 256 (6) 268 (6) 269 (6) 274 (6)
Single fixation duration (ms) 265 (7) 268 (7) 287 (8) 281 (8) 294 (8)
Gaze duration (ms) 321 (10) 338 (10) 344 (10) 347 (11) 358 (9)
Refixation probability .27 (.03) .35 (.03) .33 (.03) .32 (.03) .37 (.03)
48
Table 3
Mean (and Standard Error) Correct RT and Error Rates for the Target Word Stems in
Experiment 2
Target Prime RT (ms) Errors (%)
Prefixed Affixed 586 (3) 2.94 (0.45)
Non-affixed 587 (3) 2.74 (0.48)
Unrelated 597 (4) 4.55 (0.50)
Suffixed Affixed 602 (3) 5.08 (0.56)
Non-affixed 594 (3) 4.54 (0.46)
Unrelated 622 (2) 6.08 (0.56)
49
Figure 1
Example sentences in the boundary paradigm for (1) prefixed and (2) suffixed targets. Prior
to the reader’s eyes crossing the invisible boundary, the preview was either: (a) identical to
target, (b) same stem + different genuine affix, (c) same stem + non-affix, (d) different stem +
same affix, or (e) different stem + different affix (control).
1a. There was a feeling of mutual| mistrust between the two brothers.
1b. There was a feeling of mutual| pretrust between the two brothers.
1c. There was a feeling of mutual| whitrust between the two brothers.
1d. There was a feeling of mutual| misnomad between the two brothers.
1e. There was a feeling of mutual| outnomad between the two brothers.
2a. Students who found the examination| stressful could access help on campus.
2b. Students who found the examination| stressary could access help on campus.
2c. Students who found the examination| stressard could access help on campus.
2d. Students who found the examination| engageful could access help on campus.
2e. Students who found the examination| engageous could access help on campus.