Interhemispheric Communication via Direct Connections for Alternative Meanings of Ambiguous Words

Brain and Language 80, 77–96 (2002) doi:10.1006/brln.2001.2582, available online at http://www.idealibrary.com on

Interhemispheric Communication via Direct Connections for Alternative Meanings of Ambiguous Words Marjorie Collins Murdoch University, Perth, Western Australia, Australia

A priming experiment was used to investigate Burgess and Simpson’s (1988) claim that interhemispheric cooperation plays an essential role in the interpretation of ambiguous text. In doing so, the merits of two models of interhemispheric cooperation, the homotopic inhibition theory (Cook, 1986) and the direct connections model (Collins & Coney, 1998), were examined. Priming of alternative meanings of ambiguous words was measured using homographs and their dominant (e.g., BARK–DOG) and subordinate meanings (e.g., BARK– TREE) as related pairs in a lexical decision task, with normal university students as subjects. Stimulus pairs were temporally separated by stimulus onset asynchronies (SOAs) of 180 and 350 ms and were independently projected to the left or right visual fields (LVF or RVF). At the shorter SOA, priming was restricted to LVF–RVF presentations, with homograph primes directed to the LVF equally facilitating responses to RVF targets which were associated with their dominant and subordinate meanings. This suggests that within 180 ms, a homograph projected to the right hemisphere activates a range of alternative meanings in the left hemisphere. At an SOA of 350 ms, LVF–RVF priming was obtained along with RVF–LVF and RVF–RVF priming. Evidently at this stage of processing, an ambiguous word directed to either hemisphere activates a range of alternative meanings in the contralateral hemisphere, while RVF primes also activate subordinate, but not dominant meanings in the left hemisphere. A homograph directed to the LVF did not activate dominant or subordinate meanings within the right hemisphere at either SOA. Generally, ambiguous words directed to either hemisphere activated a more extensive array of meanings in the contralateral hemisphere than in the hemisphere to which the prime was directed. This confirms the importance of interhemispheric cooperation in generating alternate meanings of ambiguous words. Strong support was found for the direct connections model (Collins & Coney, 1998), but no support for the homotopic inhibition theory (Cook, 1986).  2002 Elsevier Science (USA) Key Words: cerebral hemispheres; lexical ambiguity; priming; dominant and subordinate meanings; interhemispheric communication; direct connections; homotopic inhibition.

INTRODUCTION

The left cerebral hemisphere has traditionally been attributed with superior language abilities, although recent investigations suggest that the right hemisphere also makes a significant contribution to the comprehension of language (e.g., Collins, 1999; Collins & Frew, 2001; (Collins) Abernethy & Coney, 1993; Brownell, 1988; Zaidel, White, Sakurai, & Banks, 1988). Evidence derived from clinical studies sugThis research was supported by an Australian Research Council grant (File number: 03-20-198-193). I thank Ms. Adele Summers for collecting the data, and Dr. Robert Kane for statistical advice. Address correspondence and reprint requests to Marjorie Collins, School of Psychology, Murdoch University, Murdoch, Perth, Western Australia, 6150. E-mail: [email protected]. 77 0093-934X/02 $35.00  2002 Elsevier Science (USA) All rights reserved.

78

MARJORIE COLLINS

gests that one area of language comprehension in which the right hemisphere plays a vital role is the interpretation of ambiguous text (e.g., Brownell, Potter, & Michelow, 1984; Weylman, Brownell, Roman, & Gardner, 1989; Kaplan, Brownell, Jacobs, & Gardner, 1990). There is also growing clinical evidence that cooperation between the cerebral hemispheres is essential to derive the full import of an ambiguous linguistic message (e.g., Brownell, 1988; Brownell, Simson, Bihrle, Potter & Gardner, 1990; Brownell et al., 1984). Burgess and Simpson (1988) have confirmed clinical observations of the right hemisphere’s contribution to the interpretation of ambiguous text in a priming study with normal subjects. They used homographs as ambiguous primes because they have more than one meaning and so are usually interpreted within context. These ambiguous primes (e.g., BANK) were presented in central vision and followed by targets associated with the prime’s dominant (e.g., MONEY) or subordinate meaning (e.g., RIVER). The targets were projected to the left (LVF) or right visual fields (RVF) either 35 or 750 ms after the onset of the prime. For the dominant meanings, Burgess and Simpson found equivalent priming for LVF and RVF target presentations at both of these stimulus onset asynchronies (SOAs). This suggests that after central presentation of a homograph, dominant meanings were activated in both hemispheres within 35 ms and remained active for at least 750 ms. In contrast, for subordinate meanings there was a different pattern of activation in each hemisphere. When targets were projected to the RVF within 35 ms, priming was obtained when these targets were associated with the subordinate meaning of the prime. However, no RVF priming was evident for these same targets when they were presented after an interval of 750 ms. Instead, responses to these targets were considerably slower (by 46 ms) when preceded by a prime pertaining to the subordinate meaning than when preceded by an unrelated prime. Hence, it seems that although dominant meanings of ambiguous words were still active in the left hemisphere at the longer SOA, subordinate meanings were inhibited. The pattern of results was quite different when targets were projected to the right hemisphere via the LVF. When a target related to the prime’s subordinate meaning was projected to the LVF, no priming was evident at the shorter SOA. However, at the longer SOA (750 ms), a 34 ms priming effect was present. This suggests that subordinate meanings took longer to become activated in the right hemisphere relative to the left hemisphere, but remained active for longer. These differing patterns of priming for alternative meanings presented to the LVF and RVF provide an insight to the relative contributions each hemisphere makes to the interpretation of ambiguous words. It appears that the left hemisphere may quickly select information pertaining to a single interpretation of ambiguous text, while the right hemisphere maintains multiple interpretations. This is consistent with Taylor’s view (1988) that the right hemisphere proposes and the left hemisphere disposes interpretations. Burgess and Simpson (1988) take this idea further in suggesting that when a subordinate meaning is required to disambiguate text, but the relevant information has been suppressed in the left hemisphere, the latter may call upon the right hemisphere to access memory information and provide an appropriate interpretation. If this is the case, interhemispheric communication must play an integral role in the interpretation of ambiguous text. The present study aims to investigate whether this is the case and to explore the mechanism underlying any such cross-hemisphere communication. To date, limited attention has been directed to the issue of interhemispheric cooperation in linguistic processing. Cook’s (1986) homotopic inhibition theory is one exception, and is particularly relevant here, as it specifies the mechanism underlying interhemispheric cooperation and makes clear predictions about the complementary processes of activation and inhibition in each hemisphere when deriving meaning

INTERHEMISPHERIC COMMUNICATION

79

from linguistic stimuli. Cook proposes, somewhat controversially, that similar information is represented in homologous areas of each hemisphere, with these areas connected through the corpus callosum. When a cortical area in one hemisphere is activated, neighboring areas within that hemisphere are inhibited. In the contralateral hemisphere, however, the homologous area is inhibited via the corpus callosum, and this acts to disinhibit ‘‘surround’’ areas. Consequently, a word directed to one hemisphere will activate closely related meanings therein, while peripheral meanings are inhibited. Simultaneously, peripheral meanings will be activated in the contralateral hemisphere while direct meanings are suppressed. Cook’s model (1986) provides a neat account of recent clinical observations of the complementary specialization of each hemisphere for linguistic processing, where the left hemisphere has been shown to predominate in extracting the literal, denotative aspects of language, while the right hemisphere appears more sensitive to peripheral and connotative meanings (e.g., Brownell et al., 1990; 1984; Tompkins, 1990; Beeman, 1993). Rodel, Cook, Regard, and Landis (1992) have also found support for this model in a study using normal subjects who were asked to judge whether word pairs were semantically related. They presented one word of each pair centrally, while the other was simultaneously presented to the RVF or LVF. When one word was directed to the LVF, they found that subjects were more likely to judge distant associates as semantically related. In contrast, words were less likely to be judged as semantically related when one word was directed to the RVF. Rodel et al. viewed these results as consistent with Cook’s proposal that left hemisphere activation of a word inhibits identical processing in the right hemisphere and thereby activates more distant associates in that hemisphere. However, Chiarello, Maxfield and Kahan (1995) question the viability of the homotopic inhibition theory as a general account of interhemispheric interaction on the basis of their failure to find any support for the model’s predictions in a study which used a priming task where homograph primes were paired with targets which were associated with their dominant and subordinate meanings. These primes were presented centrally, while targets appeared laterally 80, 130, and 200 ms after the onset of the prime. At each of these SOAs, they found equivalent priming in each hemisphere for dominant and subordinate meanings. This is contrary to the predictions of Cook’s model (1986) that a centrally presented ambiguous prime will activate dominant meanings, and thereby suppress subordinate meanings, in the left hemisphere. It is also contrary to his prediction that a converse pattern of activation and suppression will simultaneously occur in the right hemisphere, with priming of subordinate meanings and suppression of dominant meanings. In the current context, it is interesting that the pattern of priming observed in their study was also inconsistent with Burgess and Simpson’s (1988) proposal that subordinate meanings become active in the right hemisphere only after they have been suppressed in the left hemisphere. This led Chiarello et al. to conclude that left hemisphere meaning selection is not the mechanism by which subordinate meanings are activated in the right hemisphere. More generally, the findings from this study indicate that interhemispheric cooperation during linguistic processing is not achieved by the mechanisms proposed by either Cook (1986) or Burgess and Simpson (1988). The ‘‘direct connections model,’’ recently developed by Collins and Coney (1998), may more successfully account for interhemispheric communication during linguistic processing. A fundamental premise of this model, for which there is ample evidence, is that both hemispheres can make lexical decisions without the necessity for callosal relay of information, and words are primarily encoded in the hemisphere to which they are initially projected (e.g., (Collins) Abernethy & Coney, 1990; Chiarello, 1988; Zaidel et al., 1988). From this premise, the direct connections model proposes that

80

MARJORIE COLLINS

direct callosal connections exist between concepts in one hemisphere and particular related concepts in the opposite hemisphere, where interhemispheric communication at the sematic level can occur via activation spreading through these direct connections, rather than by transference of a coded form of the prime from one hemisphere to the other. Moreover, concepts directed to a particular hemisphere may activate related concepts in the contralateral hemisphere without activating those concepts within the directly stimulated hemisphere. Consequently in the priming procedure, a prime projected to one hemisphere may directly activate related meanings in the contralateral hemisphere, and this occurs independently of any activation of these meanings within the hemisphere initially receiving the prime. In this model, interhemispheric cooperation is integral to the extraction of meaning during linguistic processing. Support for the direct connections model can be garnered from studies which have independently projected prime–target pairs to the same and different hemispheres and have found cross-hemisphere facilitation without concomitant within-hemisphere facilitation. Several studies have observed this unusual pattern of priming. (Collins) Abernethy and Coney (1993) found cross-hemisphere facilitation when a prime projected to the LVF was followed by an associated RVF target, even though there was no concomitant within-hemisphere facilitation for LVF–LVF presentations of these pairs. We also found LVF–RVF facilitation without LVF–LVF facilitation for category exemplars confounded by association ((Collins) Abernethy & Coney, 1990). The findings from these two studies are consistent with a tenet of the direct connections model that the processing of a word within the right hemisphere can activate its associates in the left hemisphere independently of any activation of those associates within the right hemisphere itself. Even so, the most striking evidence in favor of the direct connections model comes from a study we designed to test the predictions of the direct connections model relative to two other spreading activation models (Collins & Coney, 1998). An unusually large priming facilitation, of 58 ms, was found for RVF–LVF presentations of low imagery primes and their concrete associates, but there was no accompanying LVF–LVF or RVF–RVF priming, with RTs for the related and baseline conditions differing by only 2 ms. The pattern of priming observed in that study ruled out the possibility that cross-hemisphere facilitation was accomplished by transference of a coded form of the prime from one hemisphere to the other. Rather, consistent with the direct connections model, the evidence was consistent with the conclusion that low imagery words projected to the left hemisphere directly activated their concrete associates in the right hemisphere without activating these associates within the left hemisphere itself. Hence, the findings from these three studies clearly demonstrate that cross-hemisphere priming can occur independently of within-hemisphere priming, with a word directed to one hemisphere directly activating related concepts in the contralateral hemisphere without activating these concepts in the hemisphere initially receiving, and semantically encoding, the prime. This raises the possibility that cross-hemisphere cooperation in the generation of alternative meanings of ambiguous text occurs in accordance with the principles of the direct connections model, where semantic encoding of ambiguous text directed to one hemisphere initiates the activation of a range of alternative meanings in the contralateral hemisphere via direct callosal connections. To elucidate the mechanism underlying cross-hemisphere cooperation in the extraction of alternative meanings during linguistic processing, the current study examined the veracity of the differing predictions of the direct connections and homotopic inhibition theories. An effective procedure to test these predictions is the priming task, where prime and target pairs are independently projected to either the RVF or LVF. Both within- and cross-hemisphere priming conditions were included in the

INTERHEMISPHERIC COMMUNICATION

81

design for this purpose. To assess the activation and suppression of alternative meanings of ambiguous words in each hemisphere, the relationship between prime–target pairs was also manipulated. Some pairs were related via the dominant meaning of homograph primes (e.g., BARK–DOG) and others were related via the prime’s subordinate meaning (e.g., BARK–TREE). As the profile of priming for alternative meanings differs in each hemisphere over time (Burgess & Simpson, 1988) the temporal interval between these pairs was manipulated. To this end, and to restrict the focus of this study to automatic processes, prime–target pairs were temporally separated by SOAs of 180 and 350 ms. The direct connections and homotopic inhibition theories both assume that the generation of meaning during linguistic processing entails true hemispheric cooperation, without duplication of processes in each hemisphere. However, the two models postulate a different mechanism by which interhemispheric communication occurs, so they make differing predictions about the pattern of priming expected under the experimental conditions of the current study. According to the homotopic inhibition theory (Cook, 1986), a prime projected to one hemisphere will activate dominant meanings therein, thereby suppressing subordinate meanings in that hemisphere. Simultaneously, subordinate meanings will be activated in the contralateral hemisphere while direct meanings are suppressed. Hence, the pattern of priming in each hemisphere will be complementary, and determined by the hemisphere initially receiving the prime, rather than hemispheric specialization per se. In contrast, the direct connections model (Collins & Coney, 1998) assumes that a prime projected to one hemisphere can directly activate alternative meanings in the contralateral hemisphere without activating these meanings in the hemisphere initially receiving the prime. Hence, priming for cross-hemisphere presentations is possible without concomitant withinhemisphere priming. As the right hemisphere appears to supply alternative meanings of ambiguous text to the left hemisphere (Burgess & Simpson, 1988), the direct connections model holds that the processing of an ambiguous prime in the right hemisphere activates alternative meanings in the left hemisphere via direct callosal connections. In this context, support for the direct connections model would be found if priming for LVF–RVF presentations occurs without any concomitant LVF–LVF priming. METHOD

Subjects Thirty-two undergraduate psychology students acted as subjects. Data for 3 of these subjects were discarded as their error rates exceeded 30% in either the word or nonword condition in the first or second experimental session. Fifteen female and 13 male subjects remained, their mean age being 22.6 years (SD ⫽ 6.08). All were predominantly right handed according to Bryden’s (1982) hand preference questionnaire (mean handedness quotient: ⫹0.67; SD ⫽ 0.25) and had normal or corrected-to-normal vision and English was their first language.

Apparatus Subjects were tested in a well-lit room containing an IBM personal computer system which controlled randomization of trial sequencing, stimulus presentation, timing, and data collection. The onset and offset of all stimuli was controlled by a program code that queried the hardware and synchronized writes to video memory such that each complete write occurred only during that part of a raster cycle when the raster scan was not utilizing the relevant memory locations. Stimuli were printed on a high-resolution, nonglare monitor in white, enlarged uppercase letters against a dark background. Screen intensity was set at a low level to minimize phosphor persistence. An adjustable chin rest was used to stabilize each subject’s head in the correct central position and distance from the screen, such that all stimuli were

82

MARJORIE COLLINS

presented with their innermost boundary, whether to the left or right of center, exactly 2° of visual angle from the central fixation. Reaction time (RT) was measured to a resolution of 1 ms via a centrally positioned microswitch response box connected to the user port of the microcomputer. Eye movements were monitored by a Sanyo video camera connected to a video monitor, which provided a magnified view of the subject’s eyes. EMH12 High Performance earmuffs were used to minimize any noise interference.

Design Reaction time (RT) was the principal dependent variable, with errors also recorded and analyzed. A lexical decision task was used, requiring subjects to discriminate words from nonwords using a GO– NOGO response procedure. Four experimental variables were manipulated: type of stimulus pair, stimulus list, SOA, and visual field of presentation. Repeated measures were taken on all variables. The first experimental variable, type of stimulus pair, incorporated a total of 1200 prime–target pairs, which were equally divided into two stimulus lists carefully matched on all relevant variables (see details below). Each list comprised related, neutral, and unrelated word pairs as well as word–nonword and neutral– nonword pairs. These lists were presented, in counterbalanced order, in two experimental sessions. The second experimental variable was stimulus list, where the related pairs in one stimulus list were associated through the homograph prime’s dominant meaning, while related pairs in the other list were associated via the prime’s subordinate meaning. The third experimental variable, visual field of presentation, comprised four levels: (i) RVF prime and target; (ii) LVF prime, RVF target; (iii) LVF prime and target; and (iv) RVF prime, LVF target. Stimuli were presented at 2° of visual angle from a central fixation and subtended a horizontal visual angle of between 2° and 6°. SOA was the fourth experimental variable, where the temporal interval between the onset of prime and target presentation was manipulated. In one experimental session the SOA was 180 ms, while in the other session the SOA was 350 ms. SOA conditions were presented in counterbalanced order. Stimulus pairs were presented in one visual field condition per experimental session, but were presented again in the second session. Selection of visual field for each pair was randomly determined both within and across experimental sessions, as was order of presentation of word pairs in each condition. Each subject was exposed to a unique distribution of pairs in each visual field condition, and a different sequence of prime–target conditions within and across testing sessions.

Stimulus Materials Two stimulus lists were presented, each comprised of 600 stimulus pairs. Of these, 100 were relatedword pairs, with 100 homographs as primes. Each prime was paired with a target associated with its dominant (e.g., BARK–TREE) and subordinate meaning (e.g., BARK–DOG) and placed into different stimulus lists. Hence, identical primes were used in each list, but the targets differed. All related pairs were drawn from homograph norms (Twilley, Dixon, Taylor, & Clark, 1994; Nelson, McEvoy, Walling, & Wheeler, 1980). The 100 unrelated word pairs were formed by repairing primes and targets from the related set, with the restriction that the resulting pairs featured no orthographic, phonemic, categorical or associative relationship (e.g., BARK–CUP). The 100 neutral pairs were formed by pairing the prime BLANK with each of the targets in the related set (e.g., BLANK–CUP). For the word–nonword pairs, 100 additional homographs were selected from the same homograph norms (Twilley et al., 1994; Nelson et al., 1980) and used as primes. These word primes were selected to match the primes in the related-word condition in relation to grammatical category, length, and word frequency [t (198) ⫽ .268; p ⫽ .789]. Each of these primes was paired with an orthographically legal, pronounceable nonword target. None of the resultant pairs were orthographically or phonologically similar. Nonwords were generated by unsystematically changing between one and three letters of the targets in the positive set. To equate the number of repetitions of these stimuli with those in the positive set, each of these additional word primes was re-paired with a different nonword target from the same list, giving 200 word–nonword pairs per list (e.g., AIR–CORMS and AIR–PELOT). Last, to form 100 neutral control pairs which mirrored the neutral condition in the positive set, each nonword target was also paired with the word BLANK (e.g., BLANK–CORMS). All pairs were phonemically and orthographically dissimilar, and none began with the same first letter. All words were nouns or adjectives, and all stimuli were between three and eight letters in length, with 95.5% between three and six letters. Mean frequency of primes in the positive set was 62.01 wpm (SD ⫽ 115.6), while primes in the negative set had a mean frequency of 57.42 wpm (SD ⫽ 126.33: Kucˇera & Francis, 1967). There was no significant difference in the frequencies for these two sets [t (198) ⫽ .268; p ⫽ .789]. Mean word frequency of targets in the dominant meanings set was 90.32 wpm (SD ⫽ 174.83), while targets in the subordinate set were 95.21 wpm (SD ⫽ 109.53: Kucera & Francis, 1967). Again, frequencies in these two sets were matched [t (198) ⫽ .237; p ⫽ .813].

INTERHEMISPHERIC COMMUNICATION

83

Procedure Prior to the first testing session, subjects were given instructions outlining their task as one of distinguishing target words from nonwords. No allusion was made to the presence of related pairs in the stimulus list, and subjects were advised that they could choose to ignore the prime. The necessity of maintaining fixation on the central cross during presentation of all stimuli was emphasized. Subjects were positioned in front of the display monitor with their heads positioned by a chin-rest directly in front of, and 60 cm distant from, the center of the screen. They were given 72 practice trials with the same structure as the experimental trials. No stimuli from the experimental set were used. Feedback relating to accuracy was provided after each practice trial. If the video monitor revealed any deviation of eyes from fixation during these trials, subjects were reminded of the importance of maintaining fixation at all times. Each trial began with a central fixation cross which remained on throughout the trial. After 600 ms, a prime word was displayed in the LVF or RVF for 120 ms. Either 50 ms or 230 ms after the prime disappeared, the target was presented in the LVF or RVF for 150 ms. The shorter interval was employed in the SOA 180 ms condition, and the longer interval in the SOA 350 ms condition. To minimize masking effects, the target appeared one line beneath the display location of the prime. The central fixation endured until the target was erased from the screen. The entire screen then remained blank for 1500 ms, during which the subject signaled a response. Randomization of trials ensured that subjects were unable to predict the visual field in which either the prime or target would appear. Subjects responded in accordance with a GO–NOGO procedure. When the target was a word, they were required to respond by simultaneously depressing two centrally positioned microswitches with both index fingers, the faster of the two responses being taken as RT for that trial. When the target was not a word, they were required to withhold their response. Subjects were permitted 1500 ms after erasure of the target to respond. Failure to respond within 1500 ms was treated as a NOGO response. Following trials in which an incorrect response was made, the word ERROR appeared directly above the central fixation for 32 ms. Subjects were permitted to rest after each block of 72 trials and were given feedback on their overall speed and accuracy for the preceding block(s). They were also encouraged to maintain an error rate of less than eight per block. Subjects completed two experimental sessions which differed in the SOA between prime and target. Both stimulus lists were presented, in counterbalanced order, in each session. Hence, each stimulus pair was presented once per experimental session and twice across the two experimental sessions, but each time in a different SOA condition. Each session was separated by a minimum of 2 days.

RESULTS

Reaction Time Analyses Statistical analyses were performed on mean correct reaction time (RT) for the positive set. A four-way ANOVA (2 ⫻ 2 ⫻ 3 ⫻ 4) was computed on SOA, stimulus list, type of stimulus pair, and visual field. The first step in interpreting these data was to ascertain whether responses differed as a function of stimulus list. After all, targets differed in each list, even though the primes were identical. So, although care was taken to ensure that the targets in each stimulus list were carefully matched, it is important to ascertain whether responses differed as a function of stimulus list, rather the relationship between prime and target. There was no such confound in the data. There was no main effect for stimulus list [F(1, 27) ⫽ 1.232; p ⫽ .277], with responses to targets in the dominant meanings list a mere 4 ms faster than responses to targets in the subordinate list. Stimulus list did not interact with SOA [F(1,27) ⫽ .41; p ⫽ .528], which indicates that the pattern of RTs to targets in the two stimulus lists was similar at the two SOAs. Nor did stimulus list interact with visual field [F(3, 81) ⫽ 1.747; p ⫽ .164] or type of stimulus pair [F(2, 54) ⫽ .051; p ⫽ .95], where RTs followed the same rank order for each visual field in the dominant and subordinate lists, and there was no difference in the pattern of responses to related, neutral, and unrelated targets in the two lists. Taken together, these results confirm that the two stimulus lists were well matched, as responses to targets did not differ as a function of stimulus list. Hence, priming effects for dominant and subordinate meanings can be interpreted without qualification.

84

MARJORIE COLLINS

In relation to the remaining variables, all three main effects were significant. Responses were significantly faster (by 24 ms) when prime–target pairs were separated by an SOA of 350 ms relative to 180 ms [F(1, 27) ⫽ 15.705; p ⬍ .001]. Apparently a performance advantage was gained when additional processing time was available to generate lexical decisions. The main effect for stimulus pair [F(2,54) ⫽ 30.45; p ⬍ .001] reflects an overall facilitation when targets were preceded by related primes (508 ms), relative to responses when these targets were preceded by neutral or unrelated primes (521 ms and 519 ms respectively). The main effect for visual field [F(3, 81) ⫽ 34.514; p ⬍ .001] reflects the faster responses when prime–target pairs were projected to the same visual field, relative to the conditions where they were projected to different visual fields (refer to Figs. 1–4). Responses were fastest when both prime and target were projected to the RVF (at 482 ms) and slowest when a RVF prime was followed by a LVF target (at 555 ms). Although my research commonly finds faster (and more accurate) responses for LVF–RVF than LVF–LVF presentations of prime–target pairs (e.g., (Collins) Abernethy & Coney, 1996; Collins & Coney, 1998), the current experiment found the converse, with faster RTs for LVF–LVF than LVF–RVF presentations (at 497 and 532 ms respectively). Even so, the visual field advantage obtained in the current experiment was consistent across both SOAs and stimulus list (and errors), which indicates that this effect was reliable. There were two two-way interactions. Although RTs were faster in all visual field conditions at the longer SOA, the interaction between SOA and visual field [F(3, 81) ⫽ 9.005; p ⬍ .001] reflects the greater benefit for crossed–visual field presentations separated by an SOA of 350 ms relative to 180 ms (35.5 ms) relative to the benefit for same-visual field presentations (13 ms). It appears that an additional 170 ms between the onset of prime and target assisted the generation of responses in all conditions. However, this benefit was greater when prime and target were projected to different visual fields. Perhaps additional time assisted the integration of stimulus information across the hemispheres. Visual field also interacted with type of stimulus pair [F(6, 162) ⫽ 3.458; p ⫽ .003]. Here, the magnitude of priming was greater for crossed- than within-visual field presentations, with facilitation averaging 20 and 7.5 ms respectively. Reaction times to targets in the neutral and unrelated conditions were similar in all visual field conditions (with an average 3 ms difference) except the LVF–RVF condition, where responses to unrelated targets were 12 ms faster than to neutral targets. SOA and visual field also entered into a three-way interaction with stimulus list [F(3, 81) ⫽ 4.182; p ⫽ .008]. There were no other three-way interactions, nor was there a four-way interaction [F(6, 162) ⫽ 1.315; p ⫽ .253]. To identify the source of the interaction between SOA, visual field and stimulus list, separate two-way ANOVAs (2 ⫻ 2) were calculated for each visual field condition, with SOA and stimulus list as the variables. There were no significant effects for the LVF–LVF condition. For the remaining three visual field conditions, there was a main effect for SOA [LVF–RVF: F(1, 27) ⫽ 18.95; p ⬍ .001; RVF–RVF: F(1, 27) ⫽ 6.5; p ⫽ .017; RVF–LVF: F(1, 27) ⫽ 19.574; p ⬍ .001]. No other effects were significant. This indicates that responses were significantly faster when prime and target were separated by the longer SOA in all visual field conditions except the LVF–LVF condition. Moreover, this RT advantage was consistent for the dominant and subordinate sets. The homotopic inhibition and direct connections theories make different predictions about the priming of dominant and subordinate meanings in each visual field condition. To examine this, related samples t tests were used to carry out planned comparisons between each related condition and its neutral baseline. Type I error was controlled by adjusting the critical p to .0469 using a modified Bonferroni test (Keppel, 1991). When a homograph prime was followed after 180 ms by a target

INTERHEMISPHERIC COMMUNICATION

85

pertaining to its dominant meaning, significant priming facilitation (of 13 ms) was present for LVF–RVF presentations [t(27) ⫽ 2.462; p ⫽ .02]. However, there was no priming for the remaining three visual field conditions [RVF–RVF: t(27) ⫽ .769; p ⫽ .449; LVF–LVF: t(27) ⫽ 1.78; p ⫽ .086; RVF–LVF: t (27) ⫽ 2.036; p ⫽ .052]. (Refer to Fig. 1.) When the SOA between these stimulus pairs was increased to 350 ms, significant priming facilitation was present when prime and target were projected to different visual fields [LVF–RVF: t (27) ⫽ 4.352; p ⬍ .001, 29 ms; and RVF– LVF: t (27) ⫽ 3.677; p ⫽ .001, 17 ms]. In contrast, there was no priming when these pairs were projected to the same visual field [RVF–RVF: t(27) ⫽ .896; p ⫽ .378; LVF–LVF: t(27) ⫽ .155; p ⫽ .878]. (See Fig. 2.) When a homograph prime was followed 180 ms later by a target pertaining to its subordinate meaning, the pattern of priming was identical to that for dominant meanings (see Fig. 3). That is, significant priming facilitation (of 23 ms) was present for LVF–RVF presentations [t(27) ⫽ 4.16; p ⬍ .001], but the response advantage for the related condition did not reach significance for the remaining three visual field conditions [RVF–RVF: t(27) ⫽ 1.999; p ⫽ .056; LVF–LVF: t (27) ⫽ 1.227; p ⫽ .23; RVF–LVF: t (27) ⫽ 1.569; p ⫽ .128]. However, when the SOA between the prime and its subordinate meaning was increased to 350 ms, significant priming facilitation was present for LVF–RVF [t (27) ⫽ 3.11; p ⫽ .004, 20 ms], RVF–LVF [t (27) ⫽ 3.651; p ⫽ .001, 25 ms] and RVF–RVF presentations [t(27) ⫽ 2.476; p ⫽ .02, 16 ms]. Nonetheless, there was no facilitation for LVF–LVF presentations at this SOA [t (27) ⫽ .148; p ⫽ .883] with responses to related pairs 1 ms slower than the neutral condition (see Fig. 4). Hence, these results illustrate that priming of alternative meanings of ambiguous words differs in each hemisphere as a function of time as well as dominance of the association between prime and target. At both SOAs, projection of a homograph to the right hemisphere facilitated responses to RVF targets which were associated with

FIG. 1. Mean correct RT for dominantly related, neutral, and unrelated word pairs presented in the left and right visual fields at an SOA of 180 ms.

86

MARJORIE COLLINS

FIG. 2. Mean correct RT for dominantly related, neutral, and unrelated word pairs presented in the left and right visual fields at an SOA of 350 ms.

FIG. 3. Mean correct RT for subordinately related, neutral, and unrelated word pairs presented in the left and right visual fields at an SOA of 180 ms.

INTERHEMISPHERIC COMMUNICATION

87

FIG. 4. Mean correct RT for subordinately related, neutral, and unrelated word pairs presented in the left and right visual fields at an SOA of 350 ms.

either the dominant or subordinate meaning of that prime. This suggests that an ambiguous word projected to the right hemisphere activates a range of meanings in the left hemisphere, and these alternative meanings are active in the left hemisphere from 180 to 350 ms. Importantly, there was no significant difference in the magnitude of facilitation for LVF–RVF presentations of dominant and subordinate meanings at either SOA 180 ms [t (27) ⫽ 1.335; p ⫽ .193, 13 ms cf. 23 ms] or 350 ms [t (27) ⫽ 1.045; p ⫽ .305, 29 ms cf. 20 ms). This indicates that each type of meaning was activated to an equal degree in this condition. While priming was restricted to LVF– RVF presentations when the prime and target were separated by an SOA of 180 ms, additional priming effects emerged when the SOA was increased to 350 ms. At this SOA, a homograph projected to the left hemisphere facilitated responses to LVF targets which were associated with the prime’s dominant and subordinate meanings. Again, there was no difference in the magnitude of facilitation for the two types of meaning in this condition [t(27) ⫽.907; p ⫽ .373; 17 ms cf 25 ms]. Unexpectedly, a homograph projected to the left hemisphere also facilitated responses to RVF targets which were associated with the prime’s subordinate, but not its dominant meaning. This suggests that when the temporal interval between prime and target is 350 ms, information pertaining to the dominant and subordinate meanings of a homograph is communicated across the hemispheres, but only subordinate meanings are active within the left hemisphere. There was no evidence in this experiment that a homograph directed to the LVF activated either dominant or subordinate meanings within the right hemisphere. Error Analyses Error rates for the positive set were also analyzed by means of a four-way ANOVA (2 ⫻ 2 ⫻ 3 ⫻ 4) on SOA, stimulus list, type of stimulus pair, and visual field.

88

MARJORIE COLLINS

Generally, the outcomes were consistent with the RT analyses. Like the RT data, the error data confirmed that the two stimulus lists were well matched. There was no main effect for stimulus list [F(1, 27) ⫽ 1.328; p ⫽ .259], with error rates for the dominant and subordinate lists differing by only 0.61%. Furthermore, stimulus list did not interact with SOA [F(1, 27) ⫽ .191; p ⫽ .665], visual field [F(3, 81) ⫽ 2.099; p ⫽ .107], or stimulus type [F(2, 54) ⫽ .523; p ⫽ .596]. This indicates that both speed and accuracy of responses were similar for both stimulus lists. Like the RT data, there were main effects for the remaining three variables. The main effect for visual field [F(3, 81) ⫽ 9.243; p ⬍ .001] reflected greater accuracy when prime and target were both projected to the same relative to different visual fields (refer to Table 1). The identical rankings for each visual field condition in both the accuracy and RT data indicates that there were no speed–accuracy trade-offs. The main effect for type of stimulus pair [F(2, 54) ⫽ 9.37; p ⬍ .001] reflected the greater accuracy in responding to the targets when they were preceded by a related prime (6.8% errors) relative to responses when these targets were preceded by unrelated or neutral primes (7.64 and 8.7% errors respectively). Last, the main effect for SOA [F(1, 27) ⫽ 46.418; p ⬍ .001] indicates that responses were faster and more accurate, by 4.9%, when prime and target were separated by an SOA of 350 ms, relative to the shorter SOA of 180 ms. Three interactions were significant. Like the RT data, the two-way interaction between SOA and visual field [F(3, 81) ⫽ 11.437; p ⬍ .001] reflected the greater benefit

TABLE 1 Percentage of Accurate Responses for Word Pairs Presented in Each Visual Field for the Dominant and Subordinate Stimulus Lists at SOAs 180 ms and 350 ms Visual field of presentation

Subordinate list SOA 180 ms Word pair relationship Related Neutral Unrelated Mean SOA 350 ms Word pair relationship Related Neutral Unrelated Mean Dominant List SOA 180 ms Word pair relationship Related Neutral Unrelated Mean SOA 350 ms Word pair relationship Related Neutral Unrelated Mean

RVF–RVF

LVF–RVF

RVF–LVF

LVF–LVF

97.1 93.9 95.9 95.6

86.6 85.4 87.3 86.4

87.7 80.2 82.3 83.4

90 93.2 93.8 92.3

97.1 96.9 97.5 97.2

96.3 94.6 93.6 94.8

93.2 89.6 93.6 92.1

93.2 95.4 93 93.9

96.1 94.8 95 95.3

91.1 82.7 87.5 87.1

87.5 83.4 85.2 85.4

93.4 94.1 92.1 93.2

96.9 95.7 95.7 96.1

95.7 92.7 94.6 94.3

94.5 94.1 95.2 94.6

94.8 94.1 95.4 94.8

89

INTERHEMISPHERIC COMMUNICATION

when prime and target appeared in different visual fields, relative to the same visual field, at the longer SOA. While there were less errors for RVF–RVF and LVF–LVF presentations at an SOA of 350 ms relative to 180 ms (1.19 and 1.55% respectively), this difference was considerably larger for LVF–RVF and RVF–LVF presentations (7.82 and 8.98% respectively). The interaction between type of stimulus pair and visual field [F(6, 162) ⫽ 4.328; p ⬍ .001] reflected the greater response advantage when related word pairs were presented in different visual fields relative to the samevisual field conditions. Unlike the RT data, these two variables also entered into a three-way interaction with stimulus list [F(6, 162) ⫽ 2.572; p ⫽ .021]. This interaction was analyzed with two separate ANOVAs (3 ⫻ 4) for the dominant and subordinate lists, with type of stimulus pair and visual field as variables. For the subordinate list, all effects were significant. The main effect for stimulus pair [F(2, 54) ⫽ 3.213; p ⫽ .048] reflected greater accuracy in response to targets preceded by related primes (92.66%) relative to neutral and unrelated primes (91.16 and 92.12% respectively). The main effect for visual field [F(11.69; p ⬍ .001] reflected greater response accuracy when prime–target pairs were projected to the same hemisphere (96.43% for RVF–RVF and 93.09% for LVF–LVF presentations) relative to pairs appearing in different hemispheres (90.62% for LVF–RVF and 87.77% for RVF–LVF). There was also an interaction between these variables [F(6, 162) ⫽ 5.026; p ⬍ .001]. Responses were more accurate for related pairs in all visual field conditions except LVF–LVF presentations, where responses to related targets were less accurate than the neutral condition by 1.8% (see Table 2). The ANOVA for the dominant list also returned all significant effects. The main effects for stimulus type [F(2, 54) ⫽ 7.165; p ⫽ .002] and visual field [F(3, 81) ⫽ 5.306; p ⫽ .002] reflected the same trends as the subordinate set: responses were more accurate when the target was preceded by a related prime (93.75%) relative to neutral and unrelated primes (91.45 and 92.59% respectively). Accuracy was also higher when prime and target were projected to the same visual field. The interaction between these two variables [F(6, 162) ⫽ 2.273; p ⫽ .039] reflects the response advantage for the related condition in all visual field conditions except the LVF–LVF condition, where accuracy was equal for related and neutral pairs. The largest priming effects were obtained for cross-visual field presentations (see Table 2). These differences are consistent with the RT data.

TABLE 2 Percentage of Accurate Responses for Word Targets in Each Visual Field for the Dominant and Subordinate Stimulus Lists Visual field of presentation

Subordinate list Word pair relationship Related Neutral Unrelated Priming Dominant list Word pair relationship Related Neutral Unrelated Priming

RVF–RVF

LVF–RVF

RVF–LVF

LVF–LVF

97.1 96.7 95.4 ⫹0.4%

91.4 90.4 90 ⫹1.0%

90.4 87.9 84.9 ⫹2.5%

91.6 93.4 94.3 ⫺1.8%

96.5 95.4 95.3 ⫹1.1%

93.4 91.1 87.7 ⫹2.3%

90.9 90.2 88.7 ⫹0.7%

94.1 93.8 94.1 0.0

90

MARJORIE COLLINS

DISCUSSION

Before interpreting these results, it is important to establish the hemispheric locus of initial processing of these stimuli. There is considerable evidence that in priming tasks using lexical decision, stimuli directly access the lexicon in the hemisphere to which they are initially projected, and are processed therein (e.g., Collins & Coney, 1998; (Collins) Abernethy & Coney, 1996; and see Chiarello, 1988, for a review of the evidence). Hardyck, Chiarello, Dronkers, and Simpson (1985) specifically investigated this question and concluded that lexical decisions occur in the hemisphere of input and do not require shared hemispheric resources. Similarly, Rodel et al. (1992) as well as (Collins) Abernethy and Coney (1990) concluded that their results are inconsistent with the idea that hemispheric differences in performance on the linguistic tasks they employed are due simply to degradation through callosal relay. Confirmation for such findings can be found in clinical studies, where it has been demonstrated that split brain patients are able to make lexical decisions for stimuli presented to either hemisphere (e.g., Zaidel, 1983). In the same vein, Zaidel et al. (1988) concluded that evidence from aphasics, commissurotomy, and normal subjects all suggest that the right hemisphere is able to make lexical decisions and that both hemispheres have a lexicon. Therefore, consistent with previous research in this area (e.g., Koivisto, 1999; Rodel et al., 1992; Chiarello, Burgess, Richards, & Pollock, 1990), it will be assumed that primes and targets were initially processed in the hemisphere to which they were projected. The results of the current experiment provide clear support for the direct connections model (Collins & Coney, 1998) and no support for the homotopic inhibition theory (Cook, 1986). According to the latter theory, the activation of dominant meanings within a given hemisphere initiates a complementary pattern of activation and suppression in homologous areas of the contralateral hemisphere. Hence, the pattern of priming is determined by the hemisphere to which the prime is directed rather than hemispheric specialization per se. Consequently, under the experimental conditions of the current experiment, priming effects arising from the projection of a prime to the left hemisphere should be identical to those initiated by right hemisphere primes. There was no support for this assumption in the current experiment. Instead, the pattern of facilitation initiated by left hemisphere primes differed from the facilitation observed after projection of these primes to the right hemisphere. This is particularly evident when prime and target were separated by an SOA of 350 ms. While ambiguous primes directed to the RVF facilitated responses to subordinate meanings projected to the RVF 350 ms later, there was no priming when these primes and their subordinate meanings were both projected to the LVF. Furthermore, a comparison of the priming effects for within- and cross-hemisphere presentations provides no support for the homotopic inhibition theory’s assumption that projection of a prime to a given hemisphere will activate dominant meanings therein and simultaneously activate subordinate meanings in the contralateral hemisphere. Rather, in the current experiment, priming of dominant meanings was exclusively associated with cross-hemisphere presentations. That is, projection of an ambiguous prime to the RVF failed to facilitate responses to dominant meanings projected to the left hemisphere after 180 or 350 ms. Nevertheless, RVF primes did facilitate responses to subordinate meanings directed to the left hemisphere after 350 ms. The activation of subordinate, but not dominant meanings in the left hemisphere after the appearance of an ambiguous prime in the RVF, is in direct contrast to the assumptions of the homotopic inhibition theory (Cook, 1986). Further contrary evidence is found in the priming for RVF–LVF presentations. The theory predicts that ambiguous primes directed to the left hemisphere will activate subordinate, but not

INTERHEMISPHERIC COMMUNICATION

91

dominant meanings in the right hemisphere. In the current experiment, a homograph projected to the RVF did facilitate responses to subordinate meanings which appeared in the LVF 350 ms later. However, equivalent RVF–LVF priming was also found for dominant meanings at this SOA. Hence, the pattern of priming obtained when homograph primes were directed to the left hemisphere is altogether inconsistent with the predictions of the homotopic inhibition theory. The homotopic inhibition theory does not fare any better when right hemisphere priming is considered. The theory assumes that the activation of dominant meanings within a given hemisphere initiates a complementary pattern of activation and suppression in homologous areas of the contralateral hemisphere. As there was no priming in the current experiment when an ambiguous prime and its dominant meaning were both projected to the LVF, simultaneous activation of subordinate meanings in the left hemisphere is precluded. Despite this, LVF primes facilitated responses to RVF targets which were associated with the subordinate meanings of these primes, and this occurred irrespective of the temporal interval separating the prime–target pairs. Furthermore, LVF primes also facilitated responses to RVF targets pertaining to the prime’s dominant meaning and to the same degree as subordinate meanings. The homotopic inhibition theory cannot account for equal activation of direct and peripheral meanings occurring simultaneously in a given hemisphere. Hence, like the findings of Chiarello et al. (1995), the pattern of priming found in the current study is inconsistent with the predictions of Cook’s homotopic inhibition theory. In contrast, the results of the current study provide strong support for the direct connections model. This support is provided by the cross-hemisphere facilitation which occurred independently of within-hemisphere facilitation in both hemispheres. First, LVF–RVF priming occurred without LVF–LVF priming at SOAs 180 and 350 ms and for both types of meaning. That is, homographs directed to the right hemisphere facilitated responses to both dominant and subordinate meanings projected to the left hemisphere 180 and 350 ms later, but did not facilitate responses to either type of meaning when they were projected directly to the right hemisphere. Hence, an ambiguous word in the LVF did not activate either its dominant or subordinate meanings within the right hemisphere. Despite this, these LVF primes activated both types of meaning in the left hemisphere. Importantly, when stimulus pairs were separated by an SOA of 180 ms, LVF–RVF facilitation occurred without concomitant facilitation for any other visual field condition. The absence of facilitation for alternative meanings projected directly to the left hemisphere at this SOA mitigates against the conclusion that the source of this priming effect was the left hemisphere and occurred subsequent to callosal transfer of LVF primes to this hemisphere. If that had been the case, priming would also be expected for RVF–RVF presentations. This did not occur for either type of meaning. Instead, this profile of priming is entirely consistent with the direct connections model, which assumes that within- and crosshemisphere facilitation are independent, as a prime projected to one hemisphere can directly activate related concepts in the opposite hemisphere via direct callosal connections and can do so without activating those words within the hemisphere initially receiving the prime. In this instance, it appears that an ambiguous word projected to the right hemisphere directly activated a range of its alternative meanings in the left hemisphere, without activating these meanings within the right hemisphere itself. LVF–RVF facilitation has also been found without LVF–LVF facilitation for category exemplars ((Collins) Abernethy & Coney, 1990) and associates ((Collins) Abernethy & Coney, 1993). Taken together, these findings indicate that a word directed to the right hemisphere can activate a range of meanings in the left hemisphere independently of the activation of these aspects of meaning in the right hemisphere itself. Further evidence consistent with the direct connections model lies in the indepen-

92

MARJORIE COLLINS

dence of cross- and within- hemisphere facilitation when the prime was directed to the left hemisphere. A homograph projected to the left hemisphere facilitated responses to dominant meanings projected to the right hemisphere 350 ms later, but did not facilitate responses to these meanings when they were projected directly to the left hemisphere. The same trend was evident when these pairs were separated by an SOA of 180 ms, with the 22 ms facilitation for RVF–LVF presentations almost reaching significance (p ⫽ .052), unlike the accompanying RVF–RVF condition, where responses to targets preceded by related and neutral primes differed by only 5 ms. In view of the absence of facilitation for LVF–LVF presentations of these pairs, and the general left hemisphere superiority in performing this task, it is unlikely that this facilitation arose through callosal transfer of the prime from left to right hemisphere (see Collins & Coney, 1998, for further discussion of this issue). Rather, this profile of priming is consistent with the direct connections model and suggests that an ambiguous word presented to the left hemisphere activates its dominant meanings in the right hemisphere via direct callosal connections, with this cross-hemisphere activation occurring independently of any activation within the left hemisphere itself. Cross-hemisphere facilitation was also evident for RVF–LVF presentations of subordinate meanings, although this was accompanied by RVF–RVF facilitation. However, facilitation for these two conditions occurred when prime–target pairs were separated by an SOA of 350 ms, but not by 180 ms. The difference in priming effects at these two intervals is consistent with previous research which indicates that the activation of various meanings differs within and between each hemisphere over time (Collins, 1999; Collins & Coney, 1998; (Collins) Abernethy & Coney, 1993, 1996; Koivisto, 1997; Nakagawa, 1991) and encapsulates the dynamic nature of information processing in the cerebral hemispheres. No firm conclusions can be made about the source of this cross-hemisphere priming for subordinate meanings separated by an SOA of 350 ms, as it was accompanied by facilitation for RVF–RVF presentations. Even so, it is unlikely that the mechanism by which the left hemisphere activates subordinate meanings in the right hemisphere differs from that identified above for dominant meanings. Hence, it seems reasonable to conclude that ambiguous primes directed to the left hemisphere activate a range of alternative meanings in the right hemisphere via direct callosal connections. This is consistent with our previous study which found RVF–LVF facilitation without RVF–RVF facilitation for low imagery primes and their concrete associates (Collins & Coney, 1998). Hence, it also appears that a word directed to the left hemisphere can activate a range of associated meanings in the right hemisphere independently of any activation within the left hemisphere itself. Generally, the present results indicate that interhemispheric cooperation plays a central role in the interpretation of ambiguous text. The predominance of facilitation effects in the cross-hemisphere conditions combined with the paucity of facilitation when stimulus pairs were projected to the same hemisphere warrant such a conclusion. The only facilitation for within-hemisphere presentations occurred when homograph primes and their subordinate meanings were both directed to the left hemisphere within 350 ms, while cross-hemisphere priming was found for all but two of the conditions examined here. Moreover, while the LVF–RVF priming confirms Burgess and Simpson’s (1988) proposal that the left hemisphere calls upon the right hemisphere to access memory information pertaining to subordinate meanings, the current results indicate that a broader conclusion is warranted. The cross-hemisphere priming effects in the current study indicate that when processing an ambiguous word, both hemispheres have the capacity to make use of memory representations of alternative meanings in the contralateral hemisphere. This conclusion is based on the obser-

INTERHEMISPHERIC COMMUNICATION

93

vation that an ambiguous word directed to the right hemisphere activated a range of alternative meanings in the left hemisphere within 180 ms, and this activation was maintained at 350 ms. Likewise, a homograph prime directed to the left hemisphere activated both dominant and subordinate meanings in the right hemisphere, although this did not manifest until the interval between prime and target was 350 ms. Importantly, an ambiguous word directed to either hemisphere activated a more extensive array of meanings in the contralateral hemisphere than in the hemisphere to which the prime was directed. Furthermore, the magnitude of facilitation for LVF–RVF presentations was equivalent for dominant and subordinate meanings at both SOAs, just as it was for RVF–LVF presentations. This suggests that meaning dominance did not mediate the level of activation of alternate meanings in the contralateral hemisphere. Rather, it appears that the processing of ambiguous text by a given hemisphere entails equivalent activation of a range of alternate meanings in the contralateral hemisphere. Although the current experiment provides support for Burgess and Simpson’s (1988) suggestion that one aspect of linguistic processing in the left hemisphere entails the activation of a range of alternative meanings in the right hemisphere, like Chiarello et al. (1995), there is no support for their proposal that subordinate meanings become active in the right hemisphere only after they have been suppressed in the left hemisphere. At an SOA of 350 ms, concurrent facilitation was found for subordinate meanings presented in the RVF–RVF and RVF–LVF conditions. Evidently, the processing of an ambiguous word by the left hemisphere involves concurrent activation of its subordinate meanings in both hemispheres. Instead, this experiment provides strong support for the view of interhemispheric communication posited by the direct connections model, where the processing of ambiguous text involves the activation of a range of alternative meanings in the contralateral hemisphere via direct callosal connections. This indicates that interhemispheric cooperation is essential for successful interpretation of ambiguous text. Therefore, an interruption of communication between the hemispheres may underlie the linguistic deficits noted in patients who have sustained damage to the right hemisphere (Brownell, 1988; Brownell et al., 1984; Weylman et al., 1989; Kaplan et al., 1990; Beeman, 1993) rather than interference in right hemisphere processing per se. The suggestion that interhemispheric cooperation is integral to the interpretation of ambiguous text is consistent with a recent study conducted by Hasbrooke and Chiarello (1998). They used a lexical decision task with centrally positioned homograph primes, and targets which were associated with the prime’s dominant or subordinate meaning. The impact of hemispheric cooperation was measured by the inclusion of bilateral and unilateral visual field presentations where, after an SOA of 750 ms, targets were projected to one visual field or redundant targets were simultaneously presented in both visual fields. The pattern of priming for unilateral and bilateral presentations indicated that the two hemispheres cooperate in processing ambiguous words, with the left hemisphere making a greater contribution. On unilateral trials, both LVF and RVF priming was found for dominant meanings, while priming for subordinate meanings was restricted to RVF presentations. Bilateral visual field trials produced priming effects which were similar to those observed for RVF presentations, although responses were generally faster and more accurate than responses to unilateral presentations. Unlike Burgess and Simpson’s (1988) study, subordinate priming was restricted to the left hemisphere. On the basis of these findings, Hasbrooke and Chiarello concluded that bilateral exhaustive activation of ambiguous word meanings is followed by a yoking of left hemisphere meaning selection with the availability of alternative meanings in the right hemisphere. The results of the current experiment indicate that this exhaustive bilateral activation of alternative

94

MARJORIE COLLINS

meanings involves cross-hemisphere activation via direct callosal connections rather than activation of the full range of meanings within a particular hemisphere. This conclusion is based upon the cross-hemisphere priming effects found at an SOA of 350 ms for LVF–RVF and RVF–LVF presentations of dominant and subordinate meanings, in contrast to within-hemisphere priming, which was restricted to RVF– RVF presentations of homograph primes and their subordinate meanings. The importance of cross-hemisphere communication in generating alternative meanings of ambiguous text has not been evident in studies which have lateralized the target, but presented the prime centrally (i.e. Chiarello et al., 1995; Burgess & Simpson, 1988). The findings of the current study differ from these studies, although this is to be expected as previous research clearly demonstrates that priming arising from central primes differs markedly from priming associated with lateral presentation of both prime and target. For instance, Chiarello et al. (1990) found symmetrical priming in each hemisphere when primes were presented centrally, but hemispheric differences emerged when primes and targets were lateralized. Similarly, Koivisto and Laine (2000) found equivalent categorical priming in each hemisphere when primes were presented centrally, while hemispheric differences emerged when both prime and target were presented laterally. One explanation for these differences is that lateral presentation stalls processing of the prime and thereby stalls the activation of associated concepts. It is also possible that a stimulus directed to one hemisphere initiates a pattern of activation that differs from the activation associated with projection of the same stimulus to the contralateral hemisphere. The differences in crossand within-hemisphere facilitation in the current experiment support this suggestion. Zaidel et al.’s (1988) conclusion that the use of central primes is problematic, as the left hemisphere may ‘‘take over in some individuals’’ (p. 87), is also consistent with this view. Their caution must also be applied when interpreting priming effects derived from a naming task. It is generally acknowledged that the naming response is generated by the left hemisphere (Koivisto, 1999). Hence, even when the focus of activation lies within the right hemisphere, the naming response is generated by the left hemisphere, which is likely to obscure the unique contributions of each hemisphere in performing this task. It is interesting in this context, that Chiarello et al. (1995) found equivalent facilitation for dominant and subordinate meanings in each hemisphere at SOAs ranging from 80 to 200 ms and used a task that involved central prime presentation and a naming response. In view of the current findings of asymmetrical priming for alternative meanings of ambiguous words, it seems likely that the combination of central primes with a naming task obscures the unique contribution of each hemisphere to the generation of alternative meanings. The differences in priming observed by Hasbrooke and Chiarello (1998) for unilateral and bilateral presentations is consistent with this suggestion, particularly as the pattern of priming in their study indicated that interhemispheric interaction plays a role in processing ambiguous word meanings. In conclusion, this study has confirmed Hasbrooke and Chiarello’s (1998) observation that interhemispheric communication plays an integral part in the interpretation of ambiguous text. It has also produced evidence consistent with the model of interhemispheric communication postulated by the direct connections model (Collins & Coney, 1998) and failed to find any support for the homotopic inhibition theory (Cook, 1986). It seems that an ambiguous word directed to either hemisphere activates a range of alternative meanings in the contralateral hemisphere, and this occurs independently of any activation of alternative meanings in the directly stimulated hemisphere. This parallels previous findings of cross-hemisphere cooperation in generating categorical and associative aspects of word meanings (Collins & Coney, 1998; (Collins) Abernethy & Coney, 1990, 1993; Collins & Frew, 2001). Taken together,

INTERHEMISPHERIC COMMUNICATION

95

these findings suggest that interhemispheric cooperation plays an essential role in the extraction of meaning in linguistic processing in general. REFERENCES Beeman, M. (1993). Semantic processing in the right hemisphere may contribute to drawing inferences from discourse. Brain and Language, 44, 80–120. Brownell, H. (1988). Appreciation of metaphoric and connotative word meanings by brain damaged patients. In C. Chiarello (Ed.), Right hemisphere contributions to lexical semantics. Berlin: SpringerVerlag. Brownell, H., Potter, H., & Michelow, D. (1984). Sensitivity to lexical denotation and connotation in brain-damaged patients: A double dissociation? Brain and Language, 22, 253–265. Brownell, H., Simson, T., Bihrle, A., Potter, H., & Gardner, H. (1990). Appreciation of metaphoric alternative word meanings by left and right brain-damaged patients. Neuropsychologia, 28, 375– 383. Bryden, M. P. (1982). Laterality: Functional asymmetry in the intact brain. New York: Academic Press. Burgess, C., & Simpson, G. (1988). Cerebral hemispheric mechanisms in the retrieval of ambiguous word meanings. Brain and Language, 33, 86–103. Chiarello, C. (1988). Lateralization of lexical processes in the normal brain: A review of visual halffield research. In H. Whitaker (Ed.), Contemporary reviews in neuropsychology. Berlin: SpringerVerlag. Chiarello, C., Burgess, C., Richards, L., & Pollock, A. (1990). Semantic and associative priming in the cerebral hemispheres: Some words do, some words don’t . . . sometimes, some places. Brain and Language, 38, 75–104. Chiarello, C., Maxfield, L., & Kahan, T. (1995). Initial right hemisphere activation of subordinate word meanings is not due to homotopic callosal inhibition. Psychonomic Bulletin & Review, 2, 375–380. Collins, M. (1999). Differences in semantic category priming in the left and right cerebral hemispheres under automatic and controlled processing conditions. Neuropsychologia, 37, 1071–1085. (Collins) Abernethy, M., & Coney, J. (1990). Semantic and phonemic priming in the cerebral hemispheres. Neuropsychologia, 28, 933–945. (Collins) Abernethy, M., & Coney, J. (1993). Associative priming in the hemispheres as a function of SOA. Neuropsychologia, 31, 1397–1409. (Collins) Abernethy, M., & Coney, J. (1996). Semantic category priming in the left cerebral hemisphere. Neuropsychologia, 34, 339–350. Collins, M., & Coney, J. (1998). Interhemispheric communication is via direct connections. Brain and Language, 64, 28–52. Collins, M., & Frew, A. (2001). Right hemisphere contributions to the comprehension of low imagery words. Laterality, 6, 111–132. Cook, N. (1986). The brain code. London: Methuen. Hardyck, C., Chiarello, C., Dronkers, N., & Simpson, G. (1985). Orienting attention within visual fields: How efficient is interhemispheric transfer? Journal of Experimental Psychology: Human Perception and Performance, 11, 650–666. Hasbrooke, R., & Chiarello, C. (1998). Bihemispheric processing of redundant bilateral lexical information. Neuropsychology, 12, 78–94. Kaplan, J., Brownell, H., Jacobs, J., & Gardner, H. (1990). The effects of right hemisphere damage on the pragmatic interpretation of conversational remarks. Brain and Language, 38, 315–333. Keppel, G. (1991). Design and analysis: A researcher’s handbook. Englewood Cliffs, NJ: Prentice-Hall. Koivisto, M. (1997). Time course of semantic activation in the cerebral hemispheres. Neuropsychologia, 35(4), 497–504. Koivisto, M. (1999). Backward priming and postlexical processing in the right hemisphere. Laterality, 3(1), 21–40. Koivisto, M., & Laine, M. (2000). Hemispheric asymmetries in activation and integration of categorical information. Laterality, 5, 1–21. Kucˇera H., & Francis, W. (1967). Computational analysis of present-day American English. Providence, RI: Brown Univ. Press.

96

MARJORIE COLLINS

Nakagawa, A. (1991). Role of anterior and posterior attention networks in hemispheric asymmetries during lexical decisions. Journal of Cognitive Neuroscience, 3, 313–321. Nelson, D., McEvoy, C., Walling, J., & Wheeler, J. (1980). The University of South Florida homograph norms. Behaviour, Research Methods and Instrumentation, 12, 16–37. Rodel, M., Cook, N., Regard, M., & Landis, T. (1992). Hemispheric dissociation in judging semantic relations: Complementarity for close and distant associates. Brain and Language, 43, 448–459. Taylor, M. (1988). The bilateral cooperative model of reading. In D. de Kerckhove & C. Lumsden (Eds.), The alphabet and the brain. Berlin: Springer-Verlag. Tompkins, C. (1990). Knowledge and strategies for processing lexical metaphor after right or left hemisphere brain damage. Journal of Speech and Hearing Research, 33, 307–316. Twilley, L., Dixon, P., Taylor, D., & Clark, K. (1994). University of Alberta norms of relative meaning frequency for 566 homographs. Memory & Cognition, 22, 111–126. Weylman, S., Brownell, H., Roman, M., & Gardner, H. (1989). Appreciation of indirect requests by left- and right-brain damaged patients: The effects of verbal context and conventionality of wording. Brain and Language, 36, 580–591. Zaidel, E. (1983). Disconnection syndrome as a model for laterality effects in the normal brain. In J. Hellige (Ed.), Cerebral hemisphere asymmetry: Method, theory, and application. New York: Praeger. Zaidel, E., White, H., Sakurai, E., & Banks, W. (1988). Hemispheric locus of lexical congruity effects: Neuropsychological reinterpretation of psycholinguistic results. In C. Chiarello (Ed.), Right hemisphere contributions to lexical semantics. Berlin: Springer-Verlag.