Theta phase and theta-gamma coupling organise the spoken language network
Curation statements for this article:-
Curated by eLife
eLife Assessment
This study provides useful data on oscillatory coupling during speech planning, particularly suggesting that theta-band phase coupling may differentiate networks involved in object recognition and semantic-lexical retrieval. However, the strength of the evidence is incomplete because the absence of a linguistic manipulation and key confounds between the naming and control conditions limit the ability to attribute the reported effects specifically to word production, leaving the main mechanistic conclusions only partially supported.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Speech production requires rapid coordination of conceptual and lexical processes across distributed cortical networks, yet the neurophysiological mechanisms enabling this coordination remain poorly understood. Oscillatory coupling has emerged as a candidate mechanism for coordinating neural activity across spatial scales. Here, we used whole-head magnetoencephalography during overt picture naming to test how phase and phase-amplitude coupling organise neural dynamics preceding articulation. We show that theta (4-8 Hz) phase coupling increases within two functionally distinct networks: a ventral occipito-temporal network supporting object recognition and a medial fronto-temporal network supporting semantic-lexical retrieval. These networks converged in the right fusiform gyrus beyond chance levels, identifying a candidate integration hub. In parallel, whole-brain analysis of theta-gamma (4-8 Hz, 40-100 Hz) phase-amplitude coupling revealed selective increases in the left inferior frontal and fusiform gyri during picture naming relative to control. Mixed-effects modelling further showed that coupling in the left fusiform correlates with trial-level response times during naming but not control trials. Together, these findings reveal the oscillatory mechanisms that implement known functional specialisation in the spoken language network; theta phase coupling coordinates distributed recognition and retrieval streams, while theta–gamma coupling modulates local computations within core word production nodes. By defining an oscillatory framework for real-time speech production, this work advances mechanistic understanding of the spoken language network and identifies frequency- and region-specific targets for neuromodulation of language production disorders.
Article activity feed
-
eLife Assessment
This study provides useful data on oscillatory coupling during speech planning, particularly suggesting that theta-band phase coupling may differentiate networks involved in object recognition and semantic-lexical retrieval. However, the strength of the evidence is incomplete because the absence of a linguistic manipulation and key confounds between the naming and control conditions limit the ability to attribute the reported effects specifically to word production, leaving the main mechanistic conclusions only partially supported.
-
Reviewer #1 (Public review):
Summary:
Akkad et al. present an MEG study of overt picture naming, designed to identify oscillatory mechanisms underlying spoken language production. Participants named line drawings of common objects or responded "no" to scrambled versions of the same images. The authors focus on the period preceding articulation and report on an increase in theta power associated with picture naming in lingual cortex and left medial prefrontal cortex. Two networks are then identified: a left lingual-seeded network involving right ventral occipito-temporal cortex, interpreted as supporting object recognition. Second, a left medial prefrontal-seeded network involving medial temporal regions (and centres in the brain) supporting semantic-lexical retrieval streams. The authors further report increased theta-gamma …
Reviewer #1 (Public review):
Summary:
Akkad et al. present an MEG study of overt picture naming, designed to identify oscillatory mechanisms underlying spoken language production. Participants named line drawings of common objects or responded "no" to scrambled versions of the same images. The authors focus on the period preceding articulation and report on an increase in theta power associated with picture naming in lingual cortex and left medial prefrontal cortex. Two networks are then identified: a left lingual-seeded network involving right ventral occipito-temporal cortex, interpreted as supporting object recognition. Second, a left medial prefrontal-seeded network involving medial temporal regions (and centres in the brain) supporting semantic-lexical retrieval streams. The authors further report increased theta-gamma phase-amplitude coupling in left fusiform and left inferior frontal regions, with stronger left fusiform PAC associated with slower naming responses.
Strengths:
The paper provides useful insight into the network-level and temporal dynamics associated with picture naming.
Weaknesses:
The results are somewhat over-interpreted, and the paper would benefit from a broader and more cautious discussion. The main concerns relate to the control condition and the interpretation of the theta-band effects. First, it is unclear whether the control condition adequately isolates the spoken-language network. Second, the reported theta-band effect appears broadband, raising the possibility that it reflects slow evoked responses rather than genuine oscillatory activity.
-
Reviewer #2 (Public review):
Summary:
In this MEG study, the authors assess the oscillatory dynamics of object naming in 27 participants. The authors observed two distinct theta-coupled networks: a ventral occipito-temporal network linked to object recognition and a medial fronto-temporal network linked to lexico-semantic retrieval. Moreover, when contrasting object naming to a scrambled image control task, they observed increased theta-gamma phase amplitude coupling in the LIFG and fusiform gyrus. The authors conclude that theta-coupling ensures the binding of recognition and retrieval processes in object naming, while the theta-gamma phase amplitude coupling ensures local computations within those networks.
Strengths:
Addressing the question 'which oscillatory dynamics govern speech planning' is interesting and timely, since not much …
Reviewer #2 (Public review):
Summary:
In this MEG study, the authors assess the oscillatory dynamics of object naming in 27 participants. The authors observed two distinct theta-coupled networks: a ventral occipito-temporal network linked to object recognition and a medial fronto-temporal network linked to lexico-semantic retrieval. Moreover, when contrasting object naming to a scrambled image control task, they observed increased theta-gamma phase amplitude coupling in the LIFG and fusiform gyrus. The authors conclude that theta-coupling ensures the binding of recognition and retrieval processes in object naming, while the theta-gamma phase amplitude coupling ensures local computations within those networks.
Strengths:
Addressing the question 'which oscillatory dynamics govern speech planning' is interesting and timely, since not much empirical evidence on this topic is currently available. Whole-brain MEG assessing phase and phase-amplitude coupling for object naming are not commonly attempted, and the study boosts a lot of power. Furthermore, the writing is concise and clear, making the paper easy to follow and pleasant to read.
Weaknesses:
My overall assessment is that the claims and conclusions of this study are stronger than the evidence supports, and this is because of several design weaknesses which I will detail below. Moreover, this study gives too one-sided a perspective on the current state-of-the-art in the neurophysiology of speech planning/object naming, and the presented conclusions can solely be understood from that narrow theoretical perspective.
(1) One of my bigger concerns is the lack of any language-related manipulations that allows to independently isolate the processing components the authors discuss. Object naming (speech planning) is a complex, multifaceted mental operation where different representations and processing levels can overlap in time and space (e.g., Indefrey, 2011; Hickok, 2012; Dell et al., 2013; Munding et al., 2016; Strijkers & Costa, 2016; Nozari & Pinet, 2020; De Zubicaray, 2023; Pickering & Strijkers, 2025). Without attempting to isolate those specific representations and/or processes and solely looking at object naming in general, the only manner to interpret the data is by relying on a priori defined and often one-to-one mappings of space and function in light of a specific theoretical perspective. This is indeed what the authors do in the current study, by linking certain brain regions they observe (e.g., more occipital for object recognition and more fronto-temporal for lexico-semantics) to the function assigned in dual-route models (note also that the authors rely on the dual-route models of comprehension, not production per se; see below). While doing so is of course legitimate, it does strongly reduce the impact of the results since they do not concern a direct empirical demonstration of the authors' conclusions, but an indirect inference based on a single theoretical framework. To give a few examples of how this weakens the conclusions: How do the authors know the fronto-temporal network is lexico-semantic and not phonological, as many other models would pose (e.g., Indefrey & Level, 2004; 2011; Hickok, 2012; 2014; De Zubicaray, 2023)? Similarly, theta-coupling in production is not uniquely related to lexico-semantics, but could also be linked to the syllabic rate (speech rate; e.g., Giraud & Poeppel, 2012; Meyer, 2018), or to working memory in production (e.g., Piai & Zheng, 2019). The occipito-temporal network may be involved in object recognition, but it may also reflect (embodied) word semantics, in particular for imageable objects as in the current study (e.g., Strijkers & Costa, 2016; Pulvermuller, 2018). In my opinion, given the lack of independent manipulations, these (and other) alternative conclusions are equally plausible, and the authors' interpretations of the data pattern are tentative at best; this is a substantial limitation in my view.
(2) The authors may object that they do have a contrast, namely the control condition. However, the control condition and the object naming condition differ in so many crucial aspects that it does not serve as a reliable control to address the above issues with interpretational strength. The control condition solely deals with low-level visual input, but not with conceptual and language-related processing, nor with motor preparation and articulation (since repeatedly saying 'no' is a flawed control for the more complex and variable naming responses required in object naming).
(3) Moreover, the timings between the control and object naming conditions are confounded: Naming and control RTs differ by 270 ms. The 1 s pre-response window therefore spans 0.08 to 1.08 s post-stimulus in naming, but −0.19 to +0.81 s in control. Hence, part of the contrast may rather reflect 'when' something was measured rather than 'what' processing differences it is supposed to reflect.
(4) As far as I can tell, the authors did not assess potential pre-articulation motor confounds in the data via ICA and/or recorded EMG activity (e.g., Vos et al., 2010; Porcaro et al., 2015; Strijkers et al., 2017; Abbasi et al., 2021). This may be relevant since motor preparatory activity is known to emerge well before actual articulation (e.g., Ries et al., 2013), and may (pre-)track articulatory speech rate in the theta range (e.g., Abbasi et al., 2023).
(5) Why are items not modelled as random intercepts/slopes? With 200 different stimuli, item variability seems relevant to include in the statistical models.
(6) I found that the authors gave a very restricted and not up-to-date literature overview on the neural basis of object naming. They mainly consider a dual route architecture. I understand that the authors don't wish (nor have to) give a complete overview of the different neurobiological models of object naming. However, the dual route model they cite is the speech processing and comprehension model (e.g., Hickok & Poeppel, 2004; 2007), which has no clear link to object naming, and they don't mention the production versions of a dual route architecture (e.g., Ueno et al., 2011; Hickok, 2012). Moreover, as mentioned also in comment 1, several other critical models (and more relevant for object naming specifically) are neglected: classical sequential object naming models (e.g., Indefrey & Levelt, 2004; Indefrey, 2011; Roelofs, 2014), interactive models (e.g., Dell et al., 2013; Dell & Chang, 2014), or more recent parallel models (e.g., Strijkers & Costa, 2016; Pickering & Costa, 2025) (for recent reviews see: e.g., Nozari & Pinet, 2020; Kerr et al., 2023; de Zubicaray, 2023). If the authors prefer to solely focus on a single class of models, that's fine (though a model that is more tightly linked to object naming would have been more relevant), but that limited scope should at the very least be acknowledged.
-
Author response:
We thank the editors and reviewers for their thoughtful and constructive assessment of our manuscript. We appreciate their recognition of the value of investigating oscillatory dynamics during overt picture naming and welcome the opportunity to strengthen the analyses and clarify the interpretation of our findings.
Functional interpretation and experimental design
We acknowledge that the contrast between picture naming and scrambled-image rejection does not independently isolate the component processes involved in spoken language production. Although the observed patterns of theta phase coupling and theta–gamma phase-amplitude coupling are consistent with the proposed involvement of visual, semantic and language-related networks, the study was designed to investigate distributed neural dynamics during overt picture …
Author response:
We thank the editors and reviewers for their thoughtful and constructive assessment of our manuscript. We appreciate their recognition of the value of investigating oscillatory dynamics during overt picture naming and welcome the opportunity to strengthen the analyses and clarify the interpretation of our findings.
Functional interpretation and experimental design
We acknowledge that the contrast between picture naming and scrambled-image rejection does not independently isolate the component processes involved in spoken language production. Although the observed patterns of theta phase coupling and theta–gamma phase-amplitude coupling are consistent with the proposed involvement of visual, semantic and language-related networks, the study was designed to investigate distributed neural dynamics during overt picture naming, rather than to establish a one-to-one correspondence between these coupling patterns and specific psycholinguistic operations.
We will therefore revise the manuscript to distinguish more clearly between the observed neurophysiological findings and their proposed functional interpretations. We will also broaden the theoretical framework to incorporate alternative accounts of the identified networks and relevant contemporary models of language production, and reconsider claims concerning the right fusiform cortex as an integration hub.
Temporal alignment and motor-related confounds
We recognise the reviewers' concerns regarding differences in response latency and articulatory demands between the naming and control conditions. Our analyses were response-locked to capture neural dynamics preceding articulation while accommodating variability in naming latency across trials and participants. We consider this alignment important for investigating language production processes, whose timing relative to stimulus onset may vary considerably.
Nevertheless, we acknowledge that differences in response latency may result in the two conditions sampling different periods relative to stimulus presentation. We will examine the potential influence of this temporal difference through complementary analyses and clarify the rationale and limitations of the response-locked approach. We will also consider the potential contribution of differences in articulatory preparation between conditions when interpreting the findings.
Interpretation of low-frequency activity
We agree that the observed low-frequency effects warrant consideration of an alternative explanation involving transient evoked activity rather than sustained theta oscillations. We will revisit the interpretation of these findings in light of the potential contribution of transient evoked activity, drawing on our existing analyses and considering additional controls where appropriate.
Statistical modelling and stimulus characteristics
We will also revisit the statistical modelling of trial-level behavioural associations, including the potential contribution of item-level variability. We recognise the reviewers' suggestion to examine psycholinguistic properties of the naming stimuli and will assess whether the available stimulus characteristics permit informative supplementary analyses, while acknowledging that the task was not designed to independently manipulate these variables.
-
-
-
Theta phase and theta-gamma coupling organise the spoken language network https://www.biorxiv.org/content/10.64898/2026.06.03.729880v1
-
Theta phase and theta-gamma coupling organise the spoken language network https://www.biorxiv.org/content/10.64898/2026.06.03.729880v1
-
-