Rovi Guides Inc v Virgin Media Ltd & Ors [2015] EWCA Civ 1214

[2015] EWCA Civ 1214Case No A3/2014/2490
IN THE COURT OF APPEAL (CIVIL DIVISION)
[2014] EWHC 1559 (Pat)
ON APPEAL FROM THE HIGH COURT OF JUSTICE
CHANCERY DIVISION
PATENTS COURT
Mr Justice Mann
Venue Royal Courts of Justice, Strand, London, WC2A 2LLDate 26/11/2015
Before
LORD JUSTICE MOORE-BICKLORD JUSTICE KITCHINLORD JUSTICE FLOYD
Between
ROVI GUIDES INC.Appellant(1) VIRGIN MEDIA LIMITEDRespondents(2) VIRGIN MEDIA PAYMENTS LIMITEDRespondent(3) TIVO INC.Respondent
Richard Meade QC and James Abrahams (instructed by Powell Gilbert LLP) for AppellantJames Mellor QC and Andrew Lykiardopoulos QC (instructed by Marks & Clerk Solicitors LLP) for RespondentsHearing Hearing date: 10 November 2015
Judgment
[1]Rovi Guides Inc The case was reported at first instance under the name Rovi Solutions Corporation and another v Virgin Media Limited and others [2014] EWHC 1559 (Pat). Rovi Guides Inc was substituted as claimant and appellant by consent, confirmed in a consent order made by Kitchin LJ on 20 July 2015. . (“Rovi”) appeals from the decision of Mann J to revoke European Patent (UK) No 0 862 833 (“the patent”) on the counterclaim of the respondents Virgin Media Limited, Virgin Media Payments Limited (together “Virgin”) and TiVo Inc in a patent infringement action. The judge gave permission to appeal on some issues but refused it on others. Rovi has now abandoned its appeal on all the issues for which it obtained permission to appeal but continues to seek permission on a single issue: whether the judge was wrong to find lack of inventive step in respect of claims 3 and 28 of the patent. The application for permission to appeal was directed to come before us, with the appeal to follow if permission was granted. Having heard all the arguments of both sides, I would grant permission.[2]On this application and appeal Rovi was represented by Mr Richard Meade QC and Mr James Abrahams. Virgin was represented by Mr James Mellor QC and Mr Andrew Lykiardopoulos QC.

The patent

[3]The patent is entitled “Interactive special events video signal navigation system.” It claims a priority date of 20 November 1995 and is concerned with the ability of a television viewer to interact with his or her television set so as to cause it to display information, for example information about a “pay-per- view” event.[4]More specifically, the specification explains that the invention is concerned with interactive video communications and with viewer-controlled selection of programming information. Known video guide hardware is said to have suffered from the disadvantage that the viewer has to wait through a scroll of the information or manually navigate to the information. When this happens the viewer can be left unaware of current and future events that are available. The invention is therefore aimed at providing a system which informs the viewer in real time of events which are available currently or will be available in the future for viewing. It does this by alerting the viewer to the existence of available information about an event by displaying a prompt or icon on the screen. This invites the viewer to interact with the system and cause the system to display information about the event.[5]The specification goes on to explain what occurs, as the judge put it, “behind the scenes”. One way in which the system operates is by employing two signals: an identification signal and an information signal. The identification signal is deployed to indicate the availability of information (in the information signal). This will enable or prompt the viewer to interact with the system. The identification signal, which is a particular focus on this appeal, may include data defining an icon or textual matter (or data enabling access to an icon or textual matter) which is displayed on the screen and which enables the viewer to be alerted to the availability of information contained in the information signal. The signal may ensure that the identification signal is displayed at pre-selected times and for pre-selected durations.[6]Each of the signals exists in two versions, the first of which is sent from the broadcaster’s end (“head-end”) to the viewer’s television, and the second of which is put up on the viewer’s screen. Although the claims are expressed in terms of a computer, the signals will normally be processed in the user’s set top box (“STB”). This is achieved in the following way: i) The first identification and information signals are received by the STB and stored. ii) At the appropriate time (which may be defined by the identification signal), the identification signal causes a prompt (an icon or text) to be displayed to the user, superimposed over the video signal that the viewer is watching. This prompt indicates that further information is available to the viewer. iii) If the user presses the appropriate button on the remote control in response to the prompt, the associated information signal is processed and the result displayed, informing the user about what is available.[7]Whilst the specification describes the use of the two signals (identification and information) in some detail, it is also at some pains to point out that the identification signal can in some circumstances be dispensed with. This can be achieved if the prompt is incorporated into the main broadcast video signal at the head end – referred to in the art as “burning in” the prompt. The alternative embodiment with no identification signal is mentioned in the general introductory part of the specification, briefly in the description of the specific embodiments and was the subject of claim 1.[8]A paragraph of the specification on which particular attention is focused (given the narrow issue in the appeal) is paragraph [12]:
“It is further contemplated that the icon 51 employed in the navigation system may be specifically correlated to the nature of the special event identified. For example, icons configured as a projector might indicate a movie event, as a tornado might indicate a weather event, as a guitar might indicate a musical event, as an open newspaper might indicate a news event, as a football helmet might indicate a sporting event and so on.”
[9]This paragraph therefore indicates that icons may be content-specific. For that purpose, it might be thought that it would be desirable for the system to arrange to have more than one icon available to it. The specification does not, however, go into any more detail. There is, for example, no detailed description of a system with access to more than one icon. As will be seen, this paragraph is deployed by Rovi to suggest that a particular integer of claim 3 should be given a specific, narrow construction.

The claims

[10]At the start of the trial the judge allowed Rovi to amend its claims. The trial then proceeded on the basis of the amended claims, and there is now no issue concerning the allowability of these amendments. In what follows I should be taken as referring to the amended set of claims which the judge included as Annex A to his judgment. Claim 1 in this set of claims is accepted by Rovi to be invalid. Claim 2 (a product claim corresponding to process claim 27) was an independent claim, and is also no longer defended against the attack on the ground of obviousness. Rovi continues to rely, however, on dependent product claim 3 and corresponding process claim 28. The appeal was argued entirely on the basis of claim 3, dependent as it is on claim 2. It is therefore sufficient to set out claims 2 and 3 with the judge’s added letters for identification purposes, but omitting the reference numerals: Claim 2 A A system interactively controlled by a TV viewer remote for superimposing identification and information characters over normal programming on a viewer's display screen comprising: B a tuner having an input for receiving TV signals in a plurality of channels and first identification and information signals, C a first output for passing a signal of any selected one of said channels and D a second output for passing said first identification and information signals; E a computer having input means for receiving any of a plurality of control signals from the TV viewer remote, F a first output means for controlling said tuner to pass said signal of said any selected one of said channels in response to a first of said plurality of control signals from the TV viewer remote, G means for receiving said first identification and information signals passed by said second output of said tuner, G1 means for storing said first identification and information signals, and H a second output means for passing a second identification signal and a second information signal, each corresponding to at least a portion of said first identification and information signals; and I means for superimposing said signals passed by said second output means of said computer over said signal of any selected one of said channels passed by said first output of said tuner to provide a display signal for input the viewer's display screen, J said computer automatically passing said second identification signal at preselected times and for predetermined time increments, without viewer input, and K passing said second information signal in response to a second of said plurality of control signals received from the TV viewer remote during said predetermined time increments. Claim 3 A A system according to claim 2, wherein said first identification signal comprises: B i. data indicating the availability of said first information signal; C ii.(a) data defining an icon or textual matter alerting a viewer to the availability of information or(b) data enabling access to such an icon or textual matter as may be already stored on said computer; and D iii. data defining said preselected times and said predetermined time increments. Claim 2 a first output for passing a signal of any selected one of said channels and means for storing said first identification and information signals, and Claim 3

A system according to claim 2, wherein said first identification signal comprises:

[11]For convenience for the discussion which follows I have added the letters “(a)” and “(b)” in bold to integer C(ii) of claim 3, to denote the two options which that integer allows. In what follows I will omit reference to the corresponding process claims, 27 and 28.

The prior art

[12]Only one item of prior art remains relevant on this appeal. It is a PCT application number WO 95/01058 filed by Apple Computer, Inc (“Florin”). Florin was published on 5 January 1995. It is a long document, but Virgin relied only on a short passage on pages 71-73 described as “Home Shopping Interface”.[13]In this passage Florin describes a dedicated home shopping channel in which each programme is dedicated to a particular shop. When the viewer is watching that channel, a blinking “select” icon is shown in a graphic overlay panel. This occurs at the start and end of each segment. The select icon is shown in Florin as a series of concentric circles resembling a bulls-eye. The icon corresponds to a button at the centre of the viewer’s remote. By pressing the select button, the user can view additional information relating to the segment being played on the TV Shop channel. For example, the viewer can bring up a list of advertisers, in which the advertiser whose commercial was playing last is highlighted.[14]Florin does not go into detail as to how the interactive functionality which it describes is achieved at the signalling level. This was the subject of expert evidence at the trial on which the judge made findings. Rovi’s expert was Mr Stephen Johnson and Virgin’s was Mr Graham Turner. Virgin’s case, advanced through Mr Turner, was that the routine implementation of the teaching of Florin using common general knowledge would lead the skilled team to construct a system which fell within the claims of the patent.

The judgment of Mann J

[15]The judge reached very clear conclusions that the skilled team implementing Florin would arrive at a system falling within claim 2 without invention. Although Rovi’s notice of appeal sought to challenge those findings, and although the judge gave permission to appeal in respect of them, Rovi no longer attacks the judge’s findings on those claims. The judge then turned to claim 3 at paragraph 60 of his judgment. He recognised that the critical additional feature of claim 3 was that provided by item (ii) in feature C. He observed that it was not always (i.e. throughout the case) clear where the dispute about this feature lay. He said: “… in final speeches it was at one point suggested that it was somehow relevant to consider the need for multiple icons. However, that point fell away …”.[16]The judge then recorded that it was not in dispute that the first option (feature C(ii)(a)) was not obvious. As the claim contained alternatives, however, the claim would still be obvious if the second alternative (feature C(ii)(b)) was obvious. The judge then recorded Rovi’s contention that, as a matter of construction, this aspect of the claim required that the data in the identification signal had to contain the equivalent of a “pointer” to the icon which was to be shown on screen, and that the evidence did not show that it was obvious to include such a pointer in the identification signal.[17]The judge then set out the following question and answer from the evidence of Virgin’s expert Mr Turner, on which particular reliance had been placed by Rovi:
"Q. I think it follows from that last answer, Mr. Turner, that none of the proposals that you have made for what to do with Florin involve a signal sent at the time of the advert change-over which includes either an icon or a pointer to an icon. That just does not happen in your suggestions about what to do with Florin. A. Well, it is an indicator of the use of an icon. There may or may not be more icons in Florin, in this situation of multiple shopping channels, but it is the existence of that message which effectively says put up an icon that causes the context of the box knowing that it needs to put up either the icon or the relevant icon. But there is not something normally in that sort of message that would say, "
Go to address this and this". Usually it is a situation of the message itself has its own context."[18]Mr Turner’s evidence was therefore that the Florin prior art suggested a system in which a message itself caused the STB to “put up an icon”, but he was not going as far as to say that Florin suggested that the message would contain an address where a relevant icon could be found.[19]The judge observed, first, that insofar as the point turned on construction, it had not received much attention in Rovi’s skeleton. This gives rise to a complaint by Rovi that the judge had ignored a lengthy note which had been supplied to the judge on day 6 of the trial, after the evidence was closed, explaining how the point had always been in issue. The judge, however, went on to record that he had to deal with the point, however much emphasis it had or had not received. The judge dealt with it as follows:
“Since it is a construction point, before turning to what Mr Turner was actually saying, it is necessary to bear in mind what the wording of the claim is. Although Mr Meade paraphrased it as referring to a "pointer", it does not actually say that. Paragraph (ii) refers to data which "enable[s] access to an icon". That no doubt includes what Mr Meade is referring to as a pointer, but it does not actually talk in terms of pointers. And it probably depends on what one means when saying "pointer". If one talks in terms of "enabling" (which is the word used) then Mr Turner's formulation in the passage of his cross-examination referred to above is capable of falling within the wording of paragraph (ii).”
[20]The judge then had regard, in considering what Mr Turner meant in his answer, to the evidence given by Mr Turner in his reports. For example in his first report at paragraph 274 he had said this about Florin:
"The Skilled Person would understand the TV Shop icon is likely to be frequently used for multiple advertisements. It is therefore most efficient to store the icon locally and simply provide data (in an "identification signal") directing the STB to place the locally stored icon on screen."
[21]The judge concluded that this paragraph was not describing a pointer, but did talk of a form of instruction to produce the icon, which was, in his view, “enabling access” to the icon as required by the claim. The judge also referred to paragraph 38 of Mr Turner’s second report, where he had said:
"(d) So, in order to create the described prompt at the start and end of each advertisement, signals must be communicated from the head-end to the STB so that the software has the necessary input(s) for scheduling the prompt to coincide with the play-out of the advertisements. This is the solution the Skilled Person would immediately understand and it would be unsurprising and predictable because it was well known to pass control messages from head-end to STB. (e) The use of such signals is also suggested when considering the nature of the further information in Florin itself. ... As the content sent from the head-end is changing, in real-time, at the start and end of each advertisement, and as the prompt is being shown at the start and end of each advertisement, it would be very odd indeed not to send a message from the head-end at the start and end of each advertisement to initialise the display of the prompt."

The judge concluded:

“These paragraphs are not directly addressing the question of the form of instruction given to or perceived by the STB, but they do strongly suggest that there is something in the signal which actually triggers the production of the prompt rather than somehow leaving the STB to infer that the prompt is needed. Again, in my view, that is "enabling". ”
[23]Having looked at how the matter was put in Mr Turner’s reports, the judge re-visited the short passage in his cross-examination which I have set out above. With this in mind he reviewed the two preceding pages of Mr Turner’s cross-examination which were concerned with how a TV channel might wish to update the channel identifier which is stored in an STB and reproduced on screen during programmes. In that context the following exchange took place:
"Q. If we can just think about this in terms of the data that will be on the box, on the set-top box, I am not saying this is exactly how it would be done or at the bit level how it would be done, but if we think of it slightly conceptually. Somewhere on the set-top box there has to be a two-column table with the channel number and the graphic appearance of the icon. That is what you need? A. Yes. In practice it would normally be a pointer to an area of storage of graphic images, but, yes. Q. I understand your point. The actual table would be the channel number plus a pointer to an address? A. Yes. Q. And the artwork would be at the address? A. Yes. Q. But conceptually you agree it is just one ---- A. Yes, it is a one-to-one mapping of channel number. It might not be channel number but logically it is the same as channel number. It is something that says this is this virtual channel or this real channel because channel numbering becomes a bit -- it is not so -- it is easy in the analog world usually because there are physical channels. When you get into digital, then channel numbering is a bit more of an abstract thing. Q. Yes. So, if there was a desire to update that, that information, all that needs to be sent is the channel number ---- A. And the image. Q. ---- and the image. That is the only signalling you need to update ---- A. Yes and that is indeed all I was trying to say here. For flexibility you would need to be able to download a new image. Q. That need not be done live or at an advert break. A. No, I was not trying to suggest that."
Q. And the artwork would be at the address? Q. But conceptually you agree it is just one ----

Q. ---- and the image. That is the only signalling you need to update ----

The judge concluded:

“What is apparent from that is that Mr Turner was being very specific about what he was referring to as a pointer – he was referring to a table in software on the machine which would actually contain a reference to a particular address in computer memory at which the information necessary to assemble the icon would be found. It is that sort of "pointer" that he was disclaiming as being part of his identification signal in the passage on which Mr Meade put such weight. He is doing no more than that. In my view there is no inconsistency between the passage relied on by Mr Meade and the passages from his reports which I have quoted and which do not deploy the same detailed concept. In my view his evidence demonstrates that he considered that the process which he had in mind fell within the concept of "enabling", and that that "enabling" arose from data in the identification signal. The contrast (if any) between what he said in cross-examination and what he had said in his reports was not put to him at this point in his cross-examination. In my view Mr Meade can only extract what he seeks from the passage relied on first by misconstruing paragraph (ii) by using the world of pointers rather than focusing on the actual wording ("enabling") and then taking Mr Turner's oral evidence out of context. In my view Mr Turner was not being inconsistent and overall his evidence demonstrates that in his view of the obvious way of implementing Florin there is some of the data in the identification signal which enables the production of the icon (in the sense that it invokes it) within paragraph (ii). I accept that evidence. ”
[25]The judge therefore concluded that the relevant integer was to be construed in the sense that data in the identification signal must trigger or invoke the icon. There was no requirement for a specific pointer or address field in the signal. It was enough if the signal itself, or some data in the signal, called for the icon to be put up on the screen. Applying that construction, claims 3 was obvious: the judge obviously accepted Mr Turner’s evidence that it would have been obvious to use the message itself as the instruction to put up the icon, or to include a specific instruction to put up the icon. It is this construction which, on this appeal, Rovi contends is too broad. Had the judge adopted Rovi’s construction, then Rovi submit that the evidence of obviousness would not have touched the claim.

Rovi’s submissions

[26]Mr Meade QC submitted that feature C (ii) of claim 3 described two different ways, (a) and (b), of telling the set top box which of a portfolio of icons the system was to use on a particular occasion. In the first, (a), the icon is actually contained in the identification signal broadcast from the head-end. This would give the head-end flexibility as to the icon which is ultimately displayed, but would have been regarded as data intensive in 1995. In the second, (b), there was a pointer to tell the set top box where in its memory it would find the appropriate icon. This would also provide flexibility, as a variety of icons could be stored locally, and the appropriate one pointed to as appropriate. This analysis lent coherence and purpose to claim integer C(ii) and was supported by paragraph [12] of the specification, with its reference to content-specific icons, which would suggest to the skilled person the use of a multiplicity of alternative icons.[27]Mr Meade also submitted that, on the judge’s construction, there was a mis-match between options (a) and (b) of feature C(ii). Option (a) undoubtedly added something to claim 2, by requiring the icon to be sent with the identification signal. On the judge’s construction of option (b), however, which included mere triggering or invoking of the icon, nothing was added to claim 2. However, on Rovi’s construction there was no such mis-match. Both options in feature C(ii) would be seen to be adding something. Although they both used extra data capacity as compared with a simple system where the identification signal itself, or a specific instruction, caused the icon to be put up, the skilled person would understand that this was for a purpose, namely to provide additional functionality in the form of content-specific icons.[28]Mr Meade submitted that the judge had focused too heavily, and almost exclusively, on the word “enabling” in the claim, which had allowed him, wrongly to substitute alternative expressions such as “triggering” or “invoking”. These alternative expressions simply meant “cause”, whereas the claim feature, when read as a whole by the skilled person, in its proper context, meant allowing the STB to do something which it could not do before. The STB would not know where to find the icon if it were one of a portfolio stored in its memory without being told by the identification signal where to find it. The claim was using “enabling” in this sense and not in the sense of mere causation or triggering.[29]Expanding his submission that the judge had failed to construe the integer as a whole, and in context, Mr Meade stressed that the signal had to contain the icon or a pointer to the icon which was “stored on the computer” (i.e. the STB). This was a way of describing a pointer to an address in the STB’s memory. That understanding would be prompted or reinforced by paragraph [12] of the specification: the skilled reader would be caused to think how the functionality described in that paragraph would be achieved, and would conclude that the sending of the icon, or a pointer to the icon in the identification signal would be the ways of achieving it.[30]Accordingly, Mr Meade submitted that although claim 2 was apt to cover an arrangement such as that made obvious by Florin, in which the icon was caused to be put up by the mere fact of the receipt of the message or a simple “put up the icon” message, option (b) of claim feature C(ii) was calling for something more. This was in accordance with what Mr Meade said would be the expectation of the skilled reader, that dependent claims normally add technical subject matter to the claims from which they depend.[31]All this, submitted Mr Meade, was in line with the purpose of the claim integer, which was to facilitate the use of multiple icons and allow the head-end to choose an appropriate, content-specific icon. The judge had lost sight of the purpose of the integer, and, in saying that the point about multiple icons “fell away” he had expressly acknowledged this.

Virgin’s submissions

[32]Mr Mellor QC supported the judge’s construction, as it accorded with the language of the claim. Rovi’s construction was wrong, firstly, because feature C does not require a pointer to a specific address in memory. The patent was not concerned with this level of detail. Secondly, Mr Mellor submitted that the reliance on multiple icons and paragraph [12] of the specification is misplaced. Multiple icons were simply an option. Although a system which used a software table to point to a particular address in the STB memory fell within the claims, so also did a signal which has data instructing the STB to render a single prompt stored on the STB. Likewise a system where the signal itself was the trigger for the appearance of the icon. The evidence on which Rovi relies to distinguish Florin relies on a particular type of pointer, namely one which uses a software table containing references to memory addresses. The claim did not require such a table.[33]Mr Mellor contended that claims 1, 2 and 3 were together seeking to claim all ways of putting the icon up on the screen. Claim 1 was directed at the burnt-in icon, whilst claims 2 and 3 were directed at the possible ways of doing it with a separate signal. He submitted that the patentee’s overall, or most general purpose was to secure wide coverage of all these methods.[34]Mr Mellor also spent some time showing us expert evidence which showed that the skilled person would think it technically unnecessary to use a pointer to a specific address if there were only a single icon. He deployed this proposition to suggest that Rovi’s construction led to a technically nonsensical result. Given that Mr Meade had accepted that a single icon system was within the claim provided that there was a pointer to a specific address, the claim would in that case require an arrangement that the skilled person would consider technically unnecessary.

Discussion

[35]There was no dispute as to the correct approach to the construction of patent claims. Mr Mellor reminded us of the proposition that it should not be assumed that it was the patentee’s intention that optional features described in the specification have been translated into requirements of the claim: see e.g. Nokia v IPCom [2009] EWHC] 3482 (Pat) at [42]. He also reminded us of the list of principles approved by the Court of Appeal (from the summary of Lewison J at first instance) in Virgin Atlantic Airways Ltd v Premium Aircraft Interiors Ltd [2009] EWCA Civ 1062 at [5]. Whilst rightly highlighting the importance of ascertaining the inventor’s purpose, those principles include at item (vi) that:
“…purpose is not the be-all and end-all. One is still at the end of the day concerned with the meaning of the language used. Hence the other extreme of the Protocol—a mere guideline—is also ruled out by Article 69 itself. It is the terms of the claims which delineate the patentee's territory.”
[36]I would add that the process of construction is seldom assisted by taking paraphrases of the words of the claim. The judge warned himself about this danger when faced with the use of the word “pointer” as a paraphrase for what was contended to be the requirement of option (b) of feature C(ii) of claim 3. The word does not appear anywhere in the claims or, I believe, in the body of the specification. The judge was right to maintain his focus on the claim language, which does not descend to this level of detail.[37]In the end this is a very short point. I have come to the conclusion that the judge’s, and thus Virgin’s, construction was the correct one. My reasons are as follows.[38]The relevant language chosen by the patentee – I set it out in full below in deference to Mr Meade’s argument that one should not focus entirely or unduly on “enabling access” – is very general. It reads:
“the identification signal comprises ….. (b) data enabling access to such an icon … as may be already stored on said computer”
[39]Mr Meade’s argument first seeks to build in to this language an element of choice: the data must enable access to the appropriate one of a portfolio of different icons stored on the computer. I am simply unable to derive this element of choice from the language. Had that been the inventor’s purpose, the draftsman would no doubt have given effect to it by the use of language such as “the appropriate one of a number of icons”. The claim language points the other way: it is simply neutral as to the availability or otherwise of additional icons. The data must enable access to whatever icon or other textual matter “as may be already stored” on the STB. This may be multiple icons, but, importantly, it could be just one.[40]The only way in which one could sensibly arrive at the conclusion that the claim required this element of choice would be to accept Rovi’s invitation to paraphrase the claim by substituting the word “pointer” for the words actually used. I accept that a pointer would be able to identify any of a series of addresses in the memory of the STB where multiple, content-specific icons could be found. There is, however, no justification whatsoever for substituting the word pointer for the entirely general words of the claim, as the judge explained. It is true that one can find references in the expert evidence to pointers, but the expert evidence is there to assist the court with relevant technical background for the purposes of construing the claim. The task of construction is one for the court alone. There can be no doubt that “pointer” is much more specific language than “data enabling access”.[41]I accept Mr Meade’s submission that the word “enable” carries a suggestion of letting the STB do that which it was not able to do or would not have done before. However I do not think that this advances his argument. Even in the situation where the signal itself merely triggers or invokes the production of a fixed icon stored on the STB, the STB was not able to and would not display the icon absent the signal. Putting this point in a computer programmer’s terms, Mr Meade suggested that, in the “mere triggering” situation, there was no “IF/THEN” instruction in the software. I do not think that this is correct. The software no doubt analyses the incoming signal and asks itself “IF” it is an identification signal, or an identification signal including the instruction “PUT UP PROMPT”, “THEN” it should put up an icon as directed by the signal. Mr Meade’s real point was that there is no second branch in the logic, to determine which of a number of possible icons should be put up. But this is to return to the element of choice, which I do not find to be present in the claim language. I reach that conclusion from considering the claim language as a whole, and in its context.[42]It is correct that a frequently deployed canon of construction is that dependent claims would normally be expected to add technical subject matter to what is claimed in the claims from which they depend. Patent attorneys would not normally be expected to go to the trouble (and sometimes expense) of adding extra claims if they were no narrower than those which precede them. However, the judge’s construction did not mean that claim 3 had the same scope as claim 2. It is now common ground that feature C(iii) was not implicit in claim 2. Feature C(i) on the other hand was implicit. In these particular circumstances I can see no reason at all why the skilled person should assume that either option in feature C(ii) was intended to add something to claim 2. In truth claim 3 sets out to list explicitly the contents of the identification signal. That is not something which is attempted in claim 2. I would not attach weight to the fact that, on the judge’s construction, feature C(ii)(b) did not itself add anything of substance to claim 2.[43]Whilst admiring its sophistication, I am not persuaded by Mr Meade’s “mis-match” submission about options (a) and (b) in feature C(ii). I do not accept that anything is to be drawn from the fact that option(a) of feature C(ii) requires extra data capacity, and therefore something different from claim 2. I think it is unrealistic to suppose that the skilled reader would reason that, because option (a) adds something to claim 2, then he should assume that option(b) is also seeking to do that as well. I accept Mr Mellor’s submission that the skilled person is more likely to assume that the patentee is seeking to cover, in explicit language, all ways he can think of arriving at a prompt on the screen.[44]The patentee’s purpose is of course a powerful guide to the construction of patent claims. One must, however, be careful to match up any purpose one derives from a reading of the specification with the claim or claim feature under discussion. This is not always straightforward. The purpose on which Mr Meade relies arises out of the description of content-specific icons in paragraph [12] of the specification. The skilled person would, so the argument runs, infer that the patentee had chosen the language of feature C(ii)(b) specifically in order to limit this claim to an arrangement capable of providing this functionality. I see no reason why the skilled person should follow this line of reasoning at all. Rather I consider that the skilled person would infer that the patentee had chosen entirely general language which was capable of encompassing such an arrangement, but was not limited to it.[45]It is true that the judge said that the point about multiple icons appeared to have fallen away. Given that claim 3 was not high on the agenda of issues at all until a very late stage in the trial, I am not surprised that he had gained that impression. Certainly, paragraph [12] of the specification does not seem to have featured strongly in the submissions or evidence until then. Moreover, it is possible that what the judge meant was that it was not contended by Rovi that claim 3 was limited to a system with multiple icons. Mr Meade accepted before us, and I believe before the judge, that a system with a pointer to the address of a unique icon would still be within claim 3. But I have reached a very firm conclusion that paragraph [12] does not help Rovi or save claim 3 from a finding of obviousness. So, if the judge indeed misspoke, it makes no difference to the outcome.[46]I do not think, therefore, that the judge overlooked or misunderstood Rovi’s arguments about the construction of claim 3. He considered and rejected them, and in my judgment he was right to do so.[47]In arriving at my conclusion I have not relied on Mr Mellor’s point that Rovi’s construction leads to a result which the skilled person would consider technically undesirable in the case of a single icon. If claim 3 in fact requires a pointer to an address in memory so as to achieve multiple icon functionality, then the fact that a pointer is not much use in the single icon situation seems beside the point. All it shows is that the skilled person is unlikely to think of it if he is focused on a document like Florin with a single icon. As I have said however, I am not persuaded that the claim is talking about a pointer at all.[48]For those reasons, and despite Mr Meade’s very able submissions, I consider that the judge’s construction of claim 3 was correct. In those circumstances some further points made by Virgin concerning whether they were afforded a proper opportunity of attacking the obviousness of the claim on the narrower construction do not need to be considered.[49]It follows, if my Lords agree, that permission to appeal will be granted, but the appeal should be dismissed.

Kitchin LJ

[50]I agree.

Moore-Bick LJ

[51]I also agree.