From action labels to semantic structure
In the previous post, I argued that the "good services are verbs" heuristic, while valuable, is incomplete. Verbs alone cannot capture the material infrastructure, institutional rules, practice configurations, and means-ends hierarchies that constitute service reality. But there is a different critique worth developing: even if we accept that verbs are central to service design, the way we currently use them is impoverished. We treat verbs as simple action labels - "book", "apply", "check", "pay" - without attending to the rich semantic structure that linguistics reveals verbs actually carry.
Charles Fillmore's case grammar and frame semantics, extended through Peter Gärdenfors's (2017) geometric treatment of meaning, offer a far richer vocabulary for describing action. This post explores what service design might learn from taking verb semantics seriously - but also asks honestly where the analogy between linguistic and service semantics breaks down.
Fillmore's case grammar: semantic roles
Traditional grammar describes sentences in terms of syntactic roles - subject, object, indirect object - but these categories are about position in a sentence, not meaning. The subject of a sentence may be an agent who acts, a patient who is acted upon, an instrument through which action occurs, or an experiencer who perceives; the syntactic role tells us nothing about which.
Fillmore's case grammar (1968) proposed a deeper level of analysis: semantic roles (or "deep cases") that capture the actual relationships between verbs and their arguments. His original set included Agent (the animate instigator of an action), Patient (the entity affected by it), Instrument (the means by which it is accomplished), Beneficiary (the entity for whose benefit it occurs), Source (the origin of movement or transfer), Goal (its destination), Locative (the spatial location), and Temporal (the time of the action). Consider the sentence "The nurse updated the patient record using the ward terminal for the consultant". Case grammar reveals the semantic structure: the nurse is Agent, the patient record is Patient, the ward terminal is Instrument, the consultant is Beneficiary, and there is an implicit Goal - some new state of the record. This is richer than "someone did something to something"; it specifies who acted, what was affected, how the action was accomplished, and for whom.
These roles also underpin the classification of denominal verbs - nouns that become verbs - as Clark and Clark (1979) demonstrated with a corpus of over 1,300 examples, grouped by the case role the parent noun plays in the implied situation.
There is a limit to acknowledge, however. Fillmore's roles were designed for the analysis of single-clause sentences describing discrete actions. Service processes are not single clauses. A hospital discharge involves dozens of actors performing hundreds of actions over days or weeks, and the same entity plays different semantic roles at different stages: the nurse is Agent when updating records, Instrument when relaying consultant decisions, and Experiencer when sensing that a patient is not ready. Applying case grammar to services therefore requires treating it not as a complete analytical framework but as a vocabulary for decomposing individual service actions - a lens that sharpens specific moments rather than capturing the whole trajectory.
Frame semantics: structured background knowledge
Fillmore later developed case grammar into frame semantics (1976, 1982), arguing that understanding a word requires access to an entire structured background of knowledge - a frame. As Gärdenfors (2017) explains, "In linguistics, Fillmore's frame semantics is the best-known use of frames". A frame is a schematised situation or event type that provides the context for understanding particular words and sentences. The verb "buy" evokes a commercial transaction frame with roles for Buyer, Seller, Goods, and Money; understanding "Sarah bought a book" requires activating this frame, even though the sentence mentions only the Buyer and Goods while the Seller, price, and payment method remain implicitly present as frame elements that could be filled in.
For service design, this suggests that verbs like "book", "apply", and "check" each evoke frames with multiple roles. "Book" evokes a scheduling frame: who makes the booking (Booker), what is being booked (Resource), the time being reserved (Slot), who controls the resource (Provider), and why the booking is needed (Purpose). "Apply" evokes a request frame: who submits (Applicant), who decides (Authority), what is sought (Benefit), what determines success (Criteria), and what supports the application (Evidence). "Check" evokes a verification frame: who checks (Checker), what is checked (Target), what it is checked against (Standard), the outcome (Result), and what follows (Consequence).
Service patterns that specify only the verb ("Book something", "Check something") leave most of the frame implicit. A richer specification would explicitly identify which frame elements are present and how they are instantiated. Whether this explicitness is always desirable is another question - there is a risk that formalising frames turns analytical insight into bureaucratic specification, producing elaborate role descriptions that are more precise than they are useful. The value of frame analysis for service design may lie less in formal notation than in the habit of asking: what roles does this action imply, and which are we overlooking?
Events as vectors in conceptual space
Gärdenfors's The Geometry of Meaning (2017) extends Fillmore's insights into the conceptual spaces framework, treating events and actions geometrically. His core proposal is that an event is represented by vectors in conceptual space: a force vector representing the cause or impetus of the event, and a result vector representing the change brought about. As he puts it, "The agent and the patient of an event model are the two most central examples of thematic roles" (Gärdenfors, 2017); the agent is associated with the force vector (they generate the force), while the patient is associated with the result vector (they undergo the change).
This connects the semantic analysis of verbs to the state space thinking developed in an earlier post. Actions are not just labels but vectors that move entities through state space. Booking an appointment means the booker's action (force vector) transforms an appointment slot from available to reserved (result vector); submitting an application means the user's action transforms an application from draft to submitted, and the user's own state from preparing to waiting.
The vector model is illuminating for these relatively simple service actions, but it is worth asking whether it generalises. Gärdenfors developed it for the cognitive semantics of physical actions - pushing, pulling, moving objects through space - where force and result have intuitive physical meaning. Extending it to abstract service actions requires treating "force" and "result" metaphorically. What is the "force vector" of submitting an application? The user clicks a button, but the actual forces involved are institutional (a system accepting data), procedural (validation rules being applied), and social (a claim being registered with an authority). The metaphor is productive - it directs attention to causation and transformation - but it should be held lightly rather than treated as literal mechanics.
Dimensions of verb meaning
Beyond semantic roles and frame structure, verbs carry information along multiple grammatical dimensions. Comrie (1976) and Palmer (2001) have developed comprehensive accounts of aspect and modality respectively; drawing on their work and wider linguistic tradition, these dimensions can be grouped into three families relevant to service design.
The temporal dimensions - tense, aspect, and adverbial time - are the most immediately useful for service status communication. Tense locates an action relative to the moment of speaking: "the patient was discharged" (past), "is being discharged" (present), "will be discharged" (future). Aspect concerns how the action is distributed across time: "the nurse checked the record" (perfective, completed) versus "the nurse was checking the record" (imperfective, ongoing) versus "the nurse has checked the record" (perfect, completed with present relevance). And adverbial time specifies when precisely: "tomorrow at 3pm", "between 2am and 4am". These distinctions matter because service interfaces constantly communicate temporal information - "Referral sent" (perfective) versus "Referral being processed" (imperfective) indicate fundamentally different states, and the distinction between "has been verified" (perfect, with present relevance) and "was verified" (simple past) signals whether the past event still bears on the current situation.
The modal dimensions - mood, modality, and polarity - are the most interesting for service governance and connect directly to the promise framework explored in the weeknotes. Mood distinguishes declarative ("the appointment is confirmed"), interrogative ("is the appointment confirmed?"), and imperative ("confirm the appointment") - services involve all three, and a button labelled "Confirm booking" is imperative while the subsequent screen showing "Booking confirmed" is declarative. Modality distinguishes what users can do (capability), may do (permission), and must do (obligation): "You can upload documents" (affordance), "You may proceed to payment" (permission granted), "You must verify your identity" (requirement). These modal distinctions are everywhere in services but rarely made explicit in service specifications. Polarity concerns whether an action is affirmed, negated, or uncertain: error states typically involve negative polarity ("Payment could not be processed"), while genuine uncertainty ("We are checking whether your application meets the criteria") represents the indeterminate.
The quality dimensions - form and manner - specify how an action is performed. The progressive form ("the system is processing requests") signals that an action is currently underway, which is critical for services where users need to know something is happening; a loading state communicates progressive aspect. Manner specifies the quality or method of action: "processed automatically" versus "handled manually", "standard" versus "expedited". These are design choices that should be explicit rather than implicit.
It is worth noting that this eight-dimension framework is a compilation rather than a received taxonomy; no single linguist proposes exactly this set for service analysis. The dimensions are drawn from different traditions - tense and aspect from Comrie (1976), modality from Palmer (2001), frame structure from Fillmore - and assembled here for service design purposes. A different selection might serve equally well; evidentiality (the linguistic marking of how the speaker knows something) seems highly relevant to services that communicate system states but is not included. The framework's value is in the habit of attending to verb dimensions rather than in the particular eight.
What richer verb analysis reveals
What would it mean to specify a service action with this richer vocabulary? Consider "book an appointment" analysed through the full framework. The action evokes a scheduling frame with roles for Booker (the user, authenticated), Resource (clinical capacity), Instrument (the booking system), and Beneficiary (the user, or possibly a dependent). The tense is future - the booking is for a future event - but the booking action itself is present. The aspect is perfective on completion: once done, the booking is done. The mood shifts from imperative (the user initiates) to declarative (the system confirms). The modality involves preconditions - the user can book (has access), the slot may be booked (is available) - and postconditions - the user must attend or cancel. The polarity may be positive (successful booking) or negative (slot no longer available). The manner is online, phone, or in-person.
This is more verbose than "Book something" - but it is also more precise about what the action involves, what conditions govern it, and what it accomplishes. Whether this precision justifies itself depends on context. For a design team trying to understand why a booking service fails in particular ways, the frame analysis might reveal that the "Resource" role is ambiguous (is the user booking a slot, a clinician, or a room?) or that the modal structure is unclear (can users cancel, or must they?). For a team building a straightforward booking interface, the analysis might add complexity without insight. The framework's usefulness is proportional to the situation's complexity and the team's need to understand why things go wrong, not just what the service does.
Connecting to broader patterns
Talmy's (1988) force dynamics, which Gärdenfors builds on, analyses the force interactions underlying verb meaning - how causation, resistance, and enabling are encoded linguistically. This is directly relevant to service design: a service that "enables" a user to do something encodes a different force dynamic from one that "requires" a user to do something, and Talmy's framework gives vocabulary for these distinctions. Levin and Rappaport Hovav (2005) identify systematic patterns in how verbs relate to argument structure - why some verbs allow certain constructions and others do not - which suggests that service action patterns are not arbitrary but follow deeper structural regularities.
Construction grammar, as developed by Goldberg (1995) and Fillmore himself, argues that grammatical constructions carry meaning independent of the words that fill them. "X caused Y to Z" is a construction with its own semantics, regardless of whether X is a nurse, a system, or a policy. For service design, this suggests that the patterns of service actions - not just individual actions - carry meaning. A "check-and-confirm" pattern, an "apply-and-wait" pattern, a "book-and-attend" pattern each has a semantic structure that transcends the particular service instance. Recognising these as constructions rather than ad hoc sequences opens the possibility of analysing service patterns for their inherent semantic properties - what roles they create, what temporal structures they impose, what modal expectations they set.
The limits of verb-centred analysis
Several honest caveats are warranted. First, the richness of linguistic verb analysis does not automatically translate into design utility. Linguists analyse verb semantics to understand how language works; designers need frameworks that help them make better services. The translation requires showing, for specific design problems, that attending to semantic roles, frame elements, or modal distinctions produces insights that simpler methods miss. The worked examples in this post are suggestive but not conclusive.
Second, there is a risk of over-formalisation. The pseudo-specification format - listing roles, preconditions, modalities - can create an illusion of rigour that obscures rather than reveals. Service situations are messy, contested, and evolving; a formal specification that freezes them into fixed roles and conditions may miss the very features that matter most. The value of case grammar for service design may lie more in the analytical habit (asking "who is the agent here? what is the instrument? what modality governs this?") than in any particular notation.
Third, verb semantics, however rich, captures only one dimension of service reality. The previous post argued that material infrastructure, institutional rules, and practice configurations are also constitutive of services; verb analysis enriches the action dimension but does not address these others. The three frames post that follows this one extends the analysis by examining how Fillmore's linguistic frames connect to Goffman's situational frames and Dorst's design frames - moving from verb semantics to situation semantics.
Finally, the relationship between this linguistic analysis and existing service description methods deserves more attention than I can give it here. Service blueprints, journey maps, and BPMN diagrams all represent service actions, but none attends to the semantic dimensions explored above. Whether augmenting these methods with linguistic precision would improve them, or merely add complexity, is an empirical question that requires testing in practice.
What this means for the verb heuristic
The "good services are verbs" heuristic is not wrong - it is shallow. Verbs are central to service meaning, but they are not simple action labels. They are frame-evoking (each verb implies a structured situation with roles), multi-dimensional (tense, aspect, mood, modality, polarity, time, and manner all contribute to meaning), vector-like (verbs represent transformations with force and result components), and domain-spanning (verbs link object domains through force and change domains).
Taking verbs seriously means taking all of this seriously - not just slapping an "-ing" on a noun and calling it user-centred. But it also means recognising that linguistic richness is a resource for design thinking, not a specification language. The service grammar project, if it is to succeed, needs to incorporate this linguistic richness while remaining honest about where the analogy between linguistic and service semantics reaches its limits. Services are indeed about actions - but actions are far more structured than we typically acknowledge, and more situated than any formal vocabulary can fully capture.
References
Clark, E.V. and Clark, H.H. (1979). When nouns surface as verbs. Language, 55(4), 767-811.
Comrie, B. (1976). Aspect. Cambridge University Press.
Fillmore, C.J. (1968). The case for case. In E. Bach and R.T. Harms (Eds.), Universals in Linguistic Theory (pp. 1-88). Holt, Rinehart and Winston.
Fillmore, C.J. (1976). Frame semantics and the nature of language. Annals of the New York Academy of Sciences, 280, 20-32.
Fillmore, C.J. (1982). Frame semantics. In Linguistics in the Morning Calm (pp. 111-137). Hanshin Publishing.
Gärdenfors, P. (2017). The Geometry of Meaning: Semantics Based on Conceptual Spaces. MIT Press.
Goldberg, A.E. (1995). Constructions: A Construction Grammar Approach to Argument Structure. University of Chicago Press.
Levin, B. and Rappaport Hovav, M. (2005). Argument Realization. Cambridge University Press.
Palmer, F.R. (2001). Mood and Modality (2nd ed.). Cambridge University Press.
Talmy, L. (1988). Force dynamics in language and cognition. Cognitive Science, 12(1), 49-100.