The Tug of War between Taxonomic and Thematic Relations in Young Toddlers

1 Introduction

1.1 Taxonomic and Thematic Relations

Recent research suggests that adult semantic memory contains two dissociable components: a taxonomic semantic system and a thematic semantic system. The taxonomic semantic system captures information about entities that are categorically related and often visually related. Thus, CAT and HORSE are taxonomically related because they belong to the same category ANIMAL and share common attributes or features such as four-legged, has-2-eyes, etc. The thematic semantic system captures information about entities that are related through their co-occurence in events. Thus, DOG and BONE are thematically related because dogs are frequently seen or depicted chewing at a bone. There are many types of thematic relations, often only constrained by the imagination. But the most commonly studied include instrumental relations, as in COMB and HAIR, and script relations, as in WINE and RESTAURANT (Moss H. E. et al. 1994; H. E. Moss et al. 1995).

Taxonomically and thematically related objects may also be associated or unassociated. Unassociated pairs of objects are defined as being rarely, if ever, produced in word association norms (such as Nelson, McEvoy, and Schreiber 2004) as associates of each other, whereas associated pairs have attested associations in such norming studies. For example, DOG and BONE are both thematically related and associated, and DOG and CAT are both taxonomically related and associated. PARTY and MUSIC are thematically related but unassociated, and PIG and HORSE are taxonomically related but unassociated. It is also possible for objects to be associated but taxonomically and thematically unrelated, as in COUCH and POTATO, PILLAR and SOCIETY or COTTAGE and CHEESE. Table 1 provides a listing of some common taxonomic categories and thematic relations.

Table 1: Cross-categorisatoin of objects: columns list objects belonging the same taxonomic category; rows list objects that are thematically related (after Mirman, Landrigan, and Britt 2017a).

Animal	Food	Instrument
Dog	Bone	Leash
Horse	Apple	Saddle
Mouse	Cheese	Trap

1.2 Evidence for Taxonomic and Thematic Memory Systems

It is well-established that taxonomically- and thematically-related words prime each other (faster reaction times) in lexical decision and naming tasks (e.g., McNamara 2005; Neely 1991; Meyer and Schvaneveldt 1971; Lucas 2000). Priming is observed if the words are only taxonomically-related or only thematically-related, though a boost in the priming effect occurs for words that are both taxonomically and thematically related (H. E. Moss et al. 1995). Thematic priming effects are longer lasting than taxonomic priming effects, i.e., priming survives longer inter-stimulus intervals (ISIs) between prime and target for thematic pairs as compared to taxonomic pairs. However, priming is rarely observed when an intervening unrelated word occurs between prime and target. Such priming effects are typically interpreted as indicating that taxonomic and thematic memory systems are structured, i.e., that the lexical-semantic representations of words in the human brain are organised according to taxonomic and thematic principles. The contrasting dynamic characteristics of thematic vs. taxonomic priming also suggest that different mechanisms may underlie these priming effects.

Event related potential (ERP) studies have also demonstrated adult sensitivity to the taxonomic and thematic relationships between words. For example, Federmeier and Kutas (1999) demonstrated a reduced N400 to a target word following a taxonomically related prime word as compared to an unrelated prime. Similarly, Mech, Kandhadai, and Federmeier (2022) reported reduced N400 signatures to target words following either a taxonomically or thematically related prime word, indicating both taxonomic and thematic relationships between words. Interestingly, this latter study reported reduced N400 effects for both left visual field (LVF) and right visual field (RVF) presentations, suggesting that both the right and left cerebral hemispheres can elicit taxonomic and thematic priming effects. Mech, Kandhadai, and Federmeier (2022) point out that thematic priming effects are stronger than taxonomic priming effects, irrespective of the side of visual field presentations.

What does it mean to suggest that there are different taxonomic and thematic semantic systems? Evidence for the separate and systematic organisation of taxonomic and thematic semantic systems requires an account of how these meaning relations are organised in the brain, and how this organisation gives rise to differences in the functioning of taxonomically or thematically related items. Different types of evidence have accumulated over the past 10–15 years suggesting that taxonomic and thematic relations between items function differently. These include neuro-psychological and neuro-imaging studies, as well as experimental behavioural studies.

1.3 Neuro-psychological and neuro-imaging studies

Schwartz et al. (2011) studied 779 naming errors across 86 aphasic patients. The authors reported that category coordinate errors (taxonomic errors), such as a picture of an apple named as grape or pear, constituted the majority of errors in this group of aphasics. However, a small minority of errors were thematic in character, such as bone named as dog or apple named as worm. The authors argued that taxonomic errors could be attributed to lesions in the left anterior temporal lobe (ATL) whereas thematic errors could be attributed to lesions to the temporoparietal junction (TPJ) (see Figure 1). Although not constituting a behavioural double dissociation between taxonomic and thematic systems, their findings point to a neuroanatomical ¹ dissociation between the 2 systems, enabling the authors to argue for a complementary semantic systems account.

Figure 1: Primary regions of the brain

Kalénine et al. (2009) reported convergent findings in an fMRI study where Ss had to match a target picture to a thematically or taxonomically related picture. Contrastive analyses identified activation in visual areas of the brain for taxonomic matches and in the inferior parietal lobule (encompassing TPJ) for thematic matches. Other studies have also highlighted the impact of lesions in the ATL on naming errors that are primarily taxonomic in character and rarely ever thematic (e.g., Lambon Ralph et al. 2001; Rogers et al. 2004).

This is how you make a margin comment. \[ y = x^2 \] Just so you know.

Note

And here is how you do callouts: note, warning, tip, important and caution

1.4 Experimental behavioural studies

When adults are asked to name a target object, thematically related distractors tend to facilitate naming whereas taxonomically-related distractors tend to inhibit naming (de Zubicaray, Hansen, and McMahon 2013). The finding suggests that taxonomically-related items compete in semantic memory whereas thematically-related items are mutually supportive. This state-of-affairs is readily understood in the context of a memory system where taxonomically-related items compete with each other for recognition and which have excitory connections to a separate system of thematically-related items.

In an eye-tracking study using the visual world paradigm (e.g., Cooper 1974; Tanenhaus et al. 1995), Ss were directed to look at an array of 4 pictures (Kalénine et al. 2012). The pictures contained foils that were either thematically-related, taxomically-related or unrelated to a spoken target item. Although the adult Ss were equally likely to fixate the thematically and taxonomically related pictures as compared to the the unrelated foils, the time courses of these fixations were different: thematic foils tended to be fixated earlier than taxonomic foils particularly when the foils were artifacts or manipulable objects. In principle, this finding could be explained within the confines of a unitary memory system where taxonomically-related items are weakly connected with each other as compared to thematically-related items. However, Kalénine et al. (2009) and Kalénine and Buxbaum (2016) have also reported that the speed of identification of taxonomically-related and thematically-related targets is reversed for animals as compared artifacts. A connection-strength type explanation would need to impose taxonomy-specific constraints (animals vs. artifacts) to account for this reversal, undermining the suggestion of a fully unitary organisation of thematic and taxonomic semantic memory systems.

Note similarity to current study.

Landrigan and Mirman (2018) argued that if thematic and taxonomic relations were encoded in distinct semantic systems, then Ss should be faster to make relatedness judgements when judgements on consecutive trials involved the same type of relation (taxonomic or thematic), than if the type of relation shifted between trials. For example, in a triads task, Ss were asked to select which of 2 words was most related} to a prompt word. In taxonomic trials the prompt word might be dog and the choice was between wolf and car. On thematic trials the prompt might be dog and the choice between bone and car. Landrigan and Mirman (2018) reported that Ss were slower to respond when their was a switch between trial types, be it taxonomic to thematic or vice versa, than when consecutive trials involved the same type of judgement. This result points to a distinction between thematic and taxonomic processes or representations.

Note

Could say something here about organisation principles of the 2 systems mentioning feature overlap vs. direct association a la Plaut.

1.5 Neuro-computational basis of the Thematic vs. Taxonomic distinction

An influential distributed network model that implements the thematic/taxonomic processing distinction is due to Plaut (1995) and Plaut and Booth (2000). The model can capture a broad range of experimental and neuropsychological findings and offers converging evidence for the theoretical foundations of this distinction (Moss H. E. et al. 1994; Plaut 1995). For example, Plaut and Booth (2000) suggest that taxonomic priming is best explained by semantic feature overlap between prime and target items in a distributed network model, whereas associative priming occurs as a result of the frequent co-occurrence of words during learning. According to Plaut and Booth (2000), the short-lived character of taxonomic priming is explained in terms of hysteresis effects in moving from one stable state of the network to another: once the feature set of the prime is fully activated, the transition to the feature set of the target will benefit from the features it has in common with the prime but will also be inhibited by the features that differ. At short stimulus onset asynchronies (SOAs), nonoverlapping features are not fully activated. Hence, overlapping features prime the transition to the target. At long SOAs (around 800 ms), nonoverlapping features of the prime are fully active and can inhibit the transition to the target (Plaut 1995; Plaut and Booth 2000). No such inhibitory effects emerge with associative priming because associative priming does not depend on feature overlap. Facilitative associative priming occurs through direct connections between lexical forms or derives from their referents sharing a complementary relationship in time or space. Other distributed network models of semantic priming implement synaptic depression mechanisms to capture SOA effects (Lerner, Bentin, and Shriki 2012; Huber and O’Reilly 2003).

The aforementioned neuro-computational models make no strong claims as to the anatomical location of the mechanisms underlying taxomomic and thematic priming. However, the neuro-psychological and neuro-imaging studies described above attempt to identify the anatomical locus of the priming effects: The anterior temporal lobe (ATL) is at the apex of the ventral visual what pathway that specialises in high-level object recognition. It is proximal to those regions of the inferior temporal lobe that specialise in the integration of shape and colour information. The parietal temporal junction (PTJ) is part of the dorsal where pathway which provides information about spatial location and action. Thus, the ATL and the PTJ each have close connectivity to different types of object representations.

As summarised above, it has been suggested that taxonomic relationships and thematic relationships between items are encoded in the anterior temporal lobe (ATL) and the parietal temporal junction (PTJ), respectively. For example, damage to the ATL results in taxonomic These anatomical regions are at opposite ends of the temporal lobe. The auditory word form area (AWFA) is conveniently located in the posterior temporal lobe, between the ATL and PTJ, and is well-positioned to provide input to both these areas.

1.5.1 Hub-and-Spoke model

Figure 2: Hub and Spoke model: single hub version

Insofar as taxonomic relations rely upon the integration of shape and colour information, and thematic relations on spatial and action information, neuro-anatomical and behavioural dissociations between taxonomic and thematic responding, are consistent with a hub-and-spoke computational model of the taxonomic and thematic semantic memory systems (e.g., Rogers et al. 2004). On this view, a taxonomic region of the hub receives stronger connections from colour and shape features encoded in the ventral visual pathway of the temporal lobe, whereas a thematic region of the hub receives stronger connections from spatial and action features in the dorsal visual pathway. Damage to connections in the taxonomic can cause a deterioration in the ability to discriminate between some colour and shape features, and results in taxonomic errors such as incorrectly naming apple as grape. Damage to connections in the thematic hub has the potential to impact appropriate binding of spatial and action features, resulting in naming errors such as bone as dog.

Using Mermaid in Quarto

%%{init: {'theme': 'default'}}%%
flowchart LR
  subgraph parentNode["Parent Node"]
    child1["Child 1"]
    child2["Child 2"]
  end
    child1 <--> child2

Source: Article Notebook

A simple representation of this neuro-computational architecture is shown in Figure 2. A single hub receives input from multiple features—the spokes of the model. The colour and shape features have stronger connections to a taxonomic region of the hub and weaker connections to a thematic region of the hub. The action and location features have stronger connections to the thematic region of the hub and weaker connections to the taxonomic region. Auditory input from the auditory word form area (AWFA) is considered to have uniform connectivity across the hub. Proponents of this type of architecture (e.g., Jefferies et al. 2020; Lambon Ralph et al. 2017) argue that the hub is located in the ATL. Damage to ATL can result in taxonomic errors. Thematic errors can be caused by inaccurate processing in this area or to semantic control errors resulting from inaccurate processing in the PTJ or prefrontal lobes.

1.5.2 Two-systems model

An alternative approach—the —suggests that there are separate taxonomic and thematic hubs (see Schwartz et al. 2011; Mirman, Landrigan, and Britt 2017b; Zhang, Mirman, and Hoffman 2023). They are both functionally and anatomically separate, as depicted in Figure 3. Colour and shape features feed into the taxonomic hub, located in the ATL. Action and location features feed into the thematic hub, located in the PTJ. Recurrent connections in the taxonomic hub are auto-associative in nature permitting pattern completion from noisy or incomplete inputs. Recurrent connections in the thematic hub are associative enabling the encoding of co-occurring events and predictions. Damage to the taxonomic and thematic hubs can lead to taxonomic and thematic errors, respectively.

Figure 3: Hub and Spoke model: dual hub version.

According to the two-systems model, hearing the word dog activates the appropriate distributed phonological representation in the AWFA. This propagates activity to the ATL and PTJ identifying the relevant features of dog in these respective cortical areas. We presume that the Anterior Temporal Lobe and the Parietal Temporal Junction are not replicates of each other. In addition to efficiency considerations, this assumption is justified on the grounds that damage to these 2 cortical areas produce different patterns of naming errors (see Schwartz et al. 2011; Mirman, Landrigan, and Britt 2017b; Zhang, Mirman, and Hoffman 2023). However, it raises the question as to the nature of the representations that correspond to the concept DOG when they are activated in the ATL and the PTJ. In other words, what is the meaning of the word dog in these 2 cortical areas. The different patterns of naming errors associated with damage to ATL and PTJ suggests that the meanings must be different.

1.5.3 Explanatory Mechanisms

As a first approximation, accepting the two-systems model, we suppose that the ATL receives perceptual input from cortical areas associated with, say, colour and shape features. In contrast, the PTJ receives functional input from cortical areas associated with motion and spatial location information. Let us further assume that both these cortical areas have a neuro-computational architecture akin to a self-organising map (SOM). SOMs have a typological organisation such that neurons which are close together on the map come to represent items that have similar features (Kohonen 2001). This means that objects that possess similar features, e.g., cat and dog, will be represented by neurons that are close together. Damage to these cortical areas can lead to errors in identification because inappropriate neurons are activated by incoming perceptual information. However, depending on the site and extent of the damage, the most likely candidates for inappropriate activation are those neurons that are close by, because damage affects neurons and the connections between them. Localised damage will result in neighbouring areas becoming relatively more activated than in the undamaged case.

If the inputs to ATL are primarily perceptual in character, such as colour and shape attributes, then the topological organisation of items represented in ATL will reflect their similarity in terms of these attributes. Dogs will be represented in a region of ATL close to that of cats insofar as they possess similar perceptual attributes. Damage to this area of ATL may lead to the inappropriate activation of cat neurons in the presence of a dog or vice versa.

If the inputs to PTJ are primarily functional in character, such as spatial location and motion attributes, then the topological organisation of the same items in PTJ will likely be different to that found in ATL because the overlap in their shared functional characteristics may well be different from that of their perceptual characteristics. For example, cats and dogs may be represented in more distinct areas of PTJ because cats move in a different fashion to dogs, as well as typically occupying in different locations. Localised damage to PTJ need not lead to inappropriate activation of cat neurons in the presence of a dog, but perhaps to the inappropriate activation of bone neurons because they are represented close by in PTJ.

1.5.4 The Meaning of Meaning

A consequence of this separation in the representation of different features of the same object is that the meaning of the label used to name an object is distributed across distinct cortical areas. This raises the question as to how representations of the same object are linked. One way that they are linked, according to this model, is through the auditory word form representation: the auditory representation of dog points at both the taxonomic and thematic meanings. But is there a concept DOG that represents both its perceptual and functional characteristics independently of the word dog? In other words, do you need to know the word dog in order to have a fully blown concept of DOG. This is reminiscent of the problem of binding the spatial and featural properties of objects that are represented separately in the What and Where visual pathways (Mishkin, Ungerleider, and Macko 1983). Received wisdom would suggest that it is possible to possess a concept without knowing its name (see Murphy (2004) pp.389 for some amusing examples involving adults as well as Merriman, Schuster, and Hager (1991) for experimental evidence with children). However, if the concept is componential (distinct regions for different types of meaning), it is conceivable that the constituent components remain isolated in the absence of language. This idea has some mileage. Let’s call it partial meaning.

According this theory of partial meanings, concepts only realise their full meaning when activated by language. Language acts as a mediating link between the partial meanings until direct links between those partial meanings are established. This could occur as a result of the frequent co-activation of the partial meanings, leading to the formation of direct connections, independently of language. Thus, barking² might come to conjure up perceptual representations associated with DOG, such as shape and colour. Formation of direct connections between anatomically separate cortical areas, such as the ATL and PTJ, might contribute to the substantial growth of white matter observed in the brain after 2 years of age (e.g., Mabbott et al. 2006). Indeed, the mediating role of language in activating disparate areas of cortex may lead to the formation of direct connections between the corresponding conceptual areas of ATL and PTJ.

The alternative approach—the hub-and-spoke model—supposes that the conceptual representations of objects are encoded entirely within the ATL (Jefferies et al. 2020; Lambon Ralph et al. 2017). The hub receives a broad range of features, e.g., shape, colour, motion, and location, contributing to the concept’s instantiation on the cortical map. On this view, there is no need to connect partial meanings of words because the sensorimotor features that define meaning converge on a single location, i.e., the ATL.

An apparent difficulty with this approach is identifying the source of thematic relationships in ATL. DOG shares no perceptual features with BONE so what mechanism leads to one concept/word priming the other for these thematically related items? Insofar as the thematic relationship between these items originated in a frequently occurring event (e.g., dogs chewing bones), then shared functional features such as spatial location and motion might serve to provide the feature overlap to encode such thematic relationships. Indeed, exactly these shared functional features are assumed to underly the encoding of thematic relationships in the two-systems model. However, in the hub-and-spoke model little additional machinery is needed to bind together representations that share the same anatomical space or hub. In contrast, the two-systems model needs additional machinery to bind together the partial meanings encoded in ATL and PTL.

The Controlled Semantic Cognition (CSC) model (Jefferies et al. 2020; Lambon Ralph et al. 2017) supposes that thematic relationships derive from processes of semantic control involved in retrieving or reconstructing an event. These processes are effortful and are likely to impose greater cognitive load on the individual. An advantage of the CSC account is that it can readily explain why thematic priming effects are larger and longer-lasting than taxonomic priming effects: the sustained effort needed to retrieve or reconstruct the appropriate event in memory likely invokes robust processing strategies that last at least as long as the time needed to activate the event in working memory. In contrast, taxonomic priming effects are considered the automatic outcome of feature overlap, imposing no additional demands on the memory system. Of course, on both the two-systems and CSC accounts, thematic priming effects might become automatic as a result of repeated event re-activations in working memory. Likewise, automatic priming of thematically related items might occur as a result of the repeated co-occurrence of objects and/or words in the auditory/visual environment, leading to a strengthening of connections between the object representations in ATL.

Warning

• Connectivity difficulties for CSC account: requires long distance spatial location and motion connections to ATL.
• Double dissociation difficulties for CSC account. Should not predict straightforward DDs.
• Why is thematic priming thought to be faster acting than taxonomic priming?

Note

Both types of architecture are agnostic as to the relative prevalence of taxonomic and thematic errors. The hub-and-spoke model predicts that damage to the hub could result in either type of error, depending on the location of the lesion in the hub. However, the integrated nature of the hub suggest that both error types should be readily observed. The two-systems model predicts that the nature of the error should depend on the location of the damage, and that thematic and taxonomic errors should be doubly-dissociable. In fact, it is commonly reported that taxonomic errors are more frequent than thematic errors (e.g., Schwartz et al. 2011; Rogers et al. 2004; Patterson, Nestor, and Rogers 2007). This asymmetry in error prevalence is not predicted by either of the architectures, as described here.

Given the current status of the neuropsychological and % experimental evidence, it is difficult to adjudicate between the two accounts of the neuroanatomical organisation of taxonomic and thematic memory systems. However, we can be reasonably confident that it makes sense to talk about distinct memory systems be % they anatomically separate (the two-systems account) or differentially distributed within a common semantic hub (hub-and-spoke account).

Note

Could say something here about organisation principles of the 2 systems mentioning feature overlap vs. direct association a la Plaut.

References

Cooper, Roger M. 1974. “The Control of Eye Fixation by the Meaning of Spoken Language: A New Methodology for the Real-Time Investigation of Speech Perception, Memory, and Language Processing.” Cognitive Psychology.

de Zubicaray, G. I., S. Hansen, and K. L. McMahon. 2013. “Differential Processing of Thematic and Categorical Conceptual Relations in Spoken Word Production.” Journal of Experimental Psychology: General 142 (1): 131.

Federmeier, K. D., and M. Kutas. 1999. “A Rose by Any Other Name: Long-Term Memory Structure and Sentence Processing.” Journal of Memory and Language 41 (4): 469–95.

Huber, D. E., and R. C. O’Reilly. 2003. “Persistence and Accommodation in Short-Term Priming and Other Perceptual Paradigms: Temporal Segregation Through Synaptic Depression.” Cognitive Science 27: 403--430.

Jefferies, E., H. Thompson, P. Cornelissen, and J. Smallwood. 2020. “The Neurocognitive Basis of Knowledge about Object Identity and Events: Dissociations Reflect Opposing Effects of Semantic Coherence and Control.” Philosophical Transactions of the Royal Society B 375 (1791): 20190300.

Kalénine, S., and L. J. Buxbaum. 2016. “Thematic Knowledge, Artifact Concepts, and the Left Posterior Temporal Lobe: Where Action and Object Semantics Converge.” Cortex 82: 164–78.

Kalénine, S., D. Mirman, E. L. Middleton, and L. J. Buxbaum. 2012. “Temporal Dynamics of Activation of Thematic and Functional Knowledge During Conceptual Processing of Manipulable Artifacts.” Journal of Experimental Psychology: Learning, Memory, and Cognition 38 (5): 1274.

Kalénine, S., C. Peyrin, C. Pichat C. Segebarth, F. Bonthoux, and M. Baciu. 2009. “The Sensory-Motor Specificity of Taxonomic and Thematic Conceptual Relations: A Behavioral and fMRI Study.” Neuroimage 44 (3): 1152–62.

Kohonen, T. 2001. Self-Organizing Maps (3rd Ed.). Springer.

Lambon Ralph, M. A., E. Jefferies, K. Patterson, and T. Rogers. 2017. “The Neural and Computational Bases of Semantic Cognition.” Nature Reviews Neuroscience 18 (1): 42–55.

Lambon Ralph, M. A., J. L McClelland, K. Patterson, C. J. Galton, and J. R. Hodges. 2001. “No Right to Speak? The Relationship Between Object Naming and Semantic Impairment: Neuropsychological Evidence and a Computational Model.” Journal of Cognitive Neuroscience 13 (3): 341–56.

Landrigan, Jon-Frederick, and Daniel Mirman. 2018. “The Cost of Switching Between Taxonomic and Thematic Semantics.” Memory & Cognition 46: 191–203.

Lerner, I., S. Bentin, and O. Shriki. 2012. “Spreading Activation in an Attractor Network with Latching Dynamics: Automatic Semantic Priming Revisited.” Cognitive Science 36 (8): 1339–82.

Lucas, M. 2000. “Semantic priming without association: A meta-analytic review.” Psychonomic Bulletin & Review 7 (4): 618–30. https://doi.org/10.3758/BF03212999.

Mabbott, Donald J, Michael Noseworthy, Eric Bouffet, Suzanne Laughlin, and Conrad Rockel. 2006. “White Matter Growth as a Mechanism of Cognitive Development in Children.” Neuroimage 33 (3): 936–46.

McNamara, Timothy P. 2005. Semantic Priming: Perspectives from Memory and Word Recognition. New York: Psychology Press.

Mech, E. N., P. Kandhadai, and K. D. Federmeier. 2022. “The Last Course of Coarse Coding: Hemispheric Similarities in Associative and Categorical Semantic Processing.” Brain and Language 229: 105123.

Merriman, William E, Joneen M Schuster, and LaurieBeth Hager. 1991. “Are Names Ever Mapped onto Preexisting Categories?” Journal of Experimental Psychology: General 120 (3): 288.

Meyer, David E, and Roger W Schvaneveldt. 1971. “Facilitation in Recognizing Pairs of Words: Evidence of a Dependence Between Retrieval Operations.” Journal of Experimental Psychology 90 (2): 227.

Mirman, D., J. F. Landrigan, and A. E. Britt. 2017b. “Taxonomic and Thematic Semantic Systems.” Psychol Bull 143 (5): 499–520.

———. 2017a. “Taxonomic and Thematic Semantic Systems.” Psychological Bulletin 143 (5): 499.

Mishkin, Mortimer, Leslie G Ungerleider, and Kathleen A Macko. 1983. “Object Vision and Spatial Vision: Two Cortical Pathways.” Trends in Neurosciences 6: 414–17.

Moss, H. E., Hare M. L., Day P., and L. K. Tyler. 1994. “A Distributed Memory Model of the Associative Boost in Semantic Priming.” Connection Science 6(4): 413–27.

Moss, Helen E, Ruth K Ostrin, Lorraine K Tyler, and William D Marslen-Wilson. 1995. “Accessing Different Types of Lexical Semantic Information: Evidence from Priming.” Journal of Experimental Psychology: Learning, Memory, and Cognition 21 (4): 863.

Murphy, Gregory. 2004. The Big Book of Concepts. MIT press.

Neely, J. H. 1991. “Semantic Priming Effects in Visual Word Recognition: A Selective Review of Current Findings and Theories.” In Basic Processes in Reading: Visual Word Recognition, edited by D. Besner and G. W. Humphreys, 264–336. Hillsdale, NJ: Erlbaum.

Nelson, D. L., C. L. McEvoy, and T. A. Schreiber. 2004. “The University of South Florida Free Association, Rhyme, and Word Fragment Norms.” Behavior Research Methods, Instruments, & Computers 36 3: 402–7.

Patterson, K., P. J. Nestor, and T. Rogers. 2007. “Where Do You Know What You Know? The Representation of Semantic Knowledge in the Human Brain.” Nature Reviews Neuroscience 8 (12): 976–87.

Plaut, D. C. 1995. “Semantic and Associative Priming in a Distributed Attractor Network.” In Proceedings of the 17th Annual Conference of the Cognitive Science Society, 17:37–42. 2. Pittsburgh, PA.

Plaut, D. C., and J. R. Booth. 2000. “Individual and Developmental Differences in Semantic Priming: Empirical and Computational Support for a Single-Mechanism Account of Lexical Processing.” Psychological Review 107: 786--823.

Rogers, T., M. A. Lambon Ralph, P. Garrard, S. Bozeat, J. L. McClelland, J. R. Hodges, and K. Patterson. 2004. “Structure and Deterioration of Semantic Memory: A Neuropsychological and Computational Investigation.” Psychological Review 111 (1): 205.

Schwartz, Myrna F, Daniel Y Kimberg, Grant M Walker, Adelyn Brecher, Olufunsho K Faseyitan, Gary S Dell, Daniel Mirman, and H Branch Coslett. 2011. “Neuroanatomical Dissociation for Taxonomic and Thematic Knowledge in the Human Brain.” Proceedings of the National Academy of Sciences 108 (20): 8520–24.

Tanenhaus, Michael K, Michael J Spivey-Knowlton, Kathleen M Eberhard, and Julie C Sedivy. 1995. “Integration of Visual and Linguistic Information in Spoken Language Comprehension.” Science 268 (5217): 1632–34.

Zhang, Y., D. Mirman, and P. Hoffman. 2023. “Taxonomic and Thematic Relations Rely on Different Types of Semantic Features: Evidence from an fMRI Meta-Analysis and a Semantic Priming Study.” Brain and Language 242: 105287.

Footnotes

1. Schwartz et al. (2011) do not claim that only TPJ lesions result in thematic errors, just that lesions in ATL and TPJ differentially impact taxonomic and thematic errors.

2. Supposing for the moment that barking is a functional attribute encoded in PTL.